WITH FRIENDS LIKE THESE, WHO NEEDS ENEMIES - RMIT ...

WITH FRIENDS LIKE THESE, WHO NEEDS ENEMIES: THE AFFORDANCES OF GAMEMECHANICS FOR THE DEVELOPMENT OF AFFECTIVE RELATIONSHIPS

A thesis submitted in fulfilment of the requirements for the degree of Doctor of Philosophy

Antimony Deor

Bachelor of Arts (The University of Melbourne)

Bachelor of Arts (Game Design) (Hons) (RMIT University)

School of Media and Communication

College of Design and Social Context

RMIT University

May 2021

I certify that except where due acknowledgement has been made, the work is that of the author alone; the work has not been submitted previously, in whole or in part, to qualify forany other academic award; the content of the thesis is the result of work which has been carried out since the official commencement date of the approved research program; any editorial work, paid or unpaid, carried out by a third party is acknowledged; and, ethics procedures and guidelines have been followed. I acknowledge the support I have received for my research through the provision of an Australian Government Research Training Program Scholarship.

Antimony Deor

20 May 2021

i

ACKNOWLEDGEMENTS

This thesis is dedicated to my grandmother, who made me promise I would acknowledge her in anything I publish.

I’d like to thank my primary supervisor Ben Byrne for being an absolute legend. He’s real goodat seeing through to the heart of things and articulating stuff perfectly. When my life had its many crisis moments, when I doubted my abilities, and when this thesis seemed completely pointless, his gentle encouragement helped me through. This thesis wouldn’t exist without him.

A thank you and an apology to my secondary supervisor Adam Nash, an actual genius who deserved a much better student than me.

I am thankful to my friend James Scott for reading whole chapters aloud to me in the final few months of editing. I got this thesis finished in time because of them.

I am indebted to my boyfriend Tim Goschnick for sticking with me through hard times, urging me to take breaks and get out of the house once in a while, and being an incredible inspiration and a true soulpal.

I am grateful to my girlfriend Tilde Joy for being patient enough to talk with me through my trickiest problems and cute enough to distract me from the rest.

The best thing that came out of this PhD was having such a thorough mental health crisis that Iqualified for experimental antidepressants. A big thank you to Aoife, Caitlyn, Sue, Jenny, Jenny, Jenny and Fenny at the Monash Alfred Psychiatric Research Centre for administering those experiments.

Lastly, I’d like to thank my cats Roast Beef, Space and Mushi for making sure I never took my keyboard for granted. Thanks guys.

ii

TABLE OF CONTENTS

Declaration............................................................................................................................................i

Acknowledgements..............................................................................................................................ii

Table of Contents................................................................................................................................iii

List of Figures.....................................................................................................................................vi

Abstract.................................................................................................................................................1

Introduction..........................................................................................................................................3

Chapter One: Special affects in games...............................................................................................12

Introduction....................................................................................................................................12Affection for Spinoza.....................................................................................................................13The power of Spinoza....................................................................................................................16What is a digital character?............................................................................................................20Affective structures of human-nonhuman relationships................................................................24

Humans and nonhumans......................................................................................................26Humans and robots...............................................................................................................32Freaks and normies...............................................................................................................36Philosophers and philosophers.............................................................................................37

Realness....................................................................................................................................38You say archetype, I say cliché, let’s call the whole thing off..................................................44

What’s your type?.................................................................................................................45Intertextuality.......................................................................................................................49

Affective characters through mechanics........................................................................................51Conclusion.....................................................................................................................................53

Chapter Two: Animals........................................................................................................................55

Introduction....................................................................................................................................55Why Play With Animals................................................................................................................57

Elves..........................................................................................................................................57Nonhuman Otters......................................................................................................................58Workhorses................................................................................................................................61Hedgehogs and Bandicoots.......................................................................................................65

Robots............................................................................................................................................67Piñatas............................................................................................................................................74Dogs...............................................................................................................................................76Chickens.........................................................................................................................................80Horses............................................................................................................................................85Rabbits...........................................................................................................................................89Cats................................................................................................................................................91

iii

Wild animals..................................................................................................................................94Conclusion.....................................................................................................................................96

Chapter Three: Enemies...................................................................................................................100

Introduction..................................................................................................................................100The enemy, a novel and successful relationship unique to computer games...............................103Enemies of philosophy.................................................................................................................106Genre............................................................................................................................................111

Fighting games........................................................................................................................111Racing games..........................................................................................................................115Stealth games...........................................................................................................................117

Enemy affects and effects............................................................................................................120Showdown...............................................................................................................................120

Enemy Spawning................................................................................................................120Player Recognising.............................................................................................................124

Fighting...................................................................................................................................127Enemy Reaction.................................................................................................................127

Penguinators..................................................................................................................130Horizon Zero Dawn.......................................................................................................134

Human Action.....................................................................................................................138Rain World.....................................................................................................................139

Dying.......................................................................................................................................142Enemy Death......................................................................................................................142Player-Character Death......................................................................................................145

Feeling.....................................................................................................................................147Nonhuman Feelings............................................................................................................147

LIM................................................................................................................................151Human Feelings..................................................................................................................153

Shadow of the Colossus.................................................................................................158Bosses..........................................................................................................................................160Conclusion...................................................................................................................................161

Chapter Four: Friends.......................................................................................................................165

Introduction..................................................................................................................................165The disappointing friends of computer games.............................................................................169Friends of philosophy..................................................................................................................172The first NPC anyone ever loved.................................................................................................176Friend affects and effects.............................................................................................................178

Co-presence.............................................................................................................................179The gaze..................................................................................................................................184Dialogue..................................................................................................................................186

Types of dialogue...............................................................................................................187Human speech actions........................................................................................................191Nonhuman speech reactions...............................................................................................196

Hot Date.........................................................................................................................201Feelings...................................................................................................................................203

Nonhuman Feelings............................................................................................................203The Last Guardian.........................................................................................................204

iv

Human Feelings..................................................................................................................208Better, faster, stronger friends......................................................................................................209

Oblivion..................................................................................................................................215Conclusion...................................................................................................................................219

Conclusion........................................................................................................................................222NPCs exist...............................................................................................................................222They affect people...................................................................................................................229Animals...................................................................................................................................231Enemies...................................................................................................................................233Friends.....................................................................................................................................239

Ludography.......................................................................................................................................243

References........................................................................................................................................252

v

LIST OF FIGURES

Fig. 1: Concept art for the game Battle Chef Brigade illustrating how each monster can be processed into different food items.

Fig. 2: The player-character approaches a Strider, a type of robot in Horizon Zero Dawn.

Fig. 3: A zoomed-in view of a chicken coop, four chickens, two ducks, and one player-character in Stardew Valley.

Fig. 4: Rabbits frolic in The Long Dark.

Fig. 5: The slugcat of Rain World in its natural habitat (lizards visible bottom left and right).

Fig. 6: The cat of Alley Cat warily watches an enemy broom.

Fig. 7: The creature of Ball, and its ball.

Fig. 8: The player moves the ball of Ball, and the creature moves to its position.

Fig. 9: In Horizon Zero Dawn, the player can see the predetermined paths of their enemies.

Fig. 10: A happy Watcher.

Fig. 11: An alert Watcher.

Fig. 12: Three attacking Watchers.

Fig. 13: A disclaimer from the The Long Dark asks players not to shoot real wolves.

Fig. 14: An abstracted model of a behaviour tree for opening a locked door..

Fig. 15: The hitbox of a Buzz Bomber.

Fig. 16: The hitbox (yellow) and hurtbox (blue) of a Crabmeat.

Fig. 17: The hitbox (pink) and hurtboxes (red) of a Caterkiller.

Fig. 18: A Penguinator attacks.

Fig. 19: The player near Final Fantasy V’s WingRaptor, who cannot yet perceive the player.

Fig. 20: Once the player-character touches the WingRaptor, a dialogue box pops up, and the player is whisked into the battle screen.

Fig. 21: The battle screen for the player’s fight with WingRaptor.

vi

Fig. 22: An arrangement of squares spelling “LIM” .

Fig. 23: The player-character square (brown) moves into view of an enemy square (blue), and is attacked.

Fig. 24: The player-character fixes itself blue, and the camera zooms in on it.

Fig. 25: A player-character wonders why everyone wants to fight her all the time.

Fig. 26: The characters of Spider stare at each other.

Fig. 27: Emily Short’s example of a typical branching dialogue tree structure.

Fig. 28: The player can cheer or heckle Aldridge T Abbington in Red Dead Redemption 2.

Fig. 29: The persuasive speech-craft mini-game from The Elder Scrolls IV: Oblivion.

Fig. 30: A nonplayer character in Hades recognises the player-character, admiring them for playing the game repeatedly.

Fig. 31: The player insults a stranger for no reason in Red Dead Redemption 2.

Fig. 32: An example of a dialogue interface from The Elder Scrolls IV: Oblivion.

Fig. 33: The interlocutor of Hot Date is unimpressed.

Note: some screenshots include a small cat shimeji (a kind of desktop digital pet), referenced as Neko in Ludography.

vii

ABSTRACT

Nonplayer characters are some of the hardest game elements to get right. A lot is demanded of

these otherworldly counterparts. From their origin as stand-ins for absent human players,

nonplayer characters have become distinguished as their own entities, and codified into many

distinct categories. Excellent progress has been made in terms of the resolution of their

graphics, the subtlety of their movement and the breadth of their perception. The diversity of

humans and relationships represented by nonplayer characters has also improved immensely

in recent years. However, there is still work to be done.

Nonplayer characters have been studied through the disciplines of game design, psychology,

AI and robotics, and media and cultural studies. These investigations touch on the emotional

impact nonplayer characters can have on human players, but mostly focus on the perceived

humanness or believability of nonplayer characters as mechanisms for affect. However, even

when a player is not immersed in a game or finds a nonplayer character unrealistic, they can

still have an important experience. This thesis addresses the inherent non-humanness of

nonplayer characters and focuses on one of the persistent problems with nonplayer

characters: their affective capacity, or their ability to modify the player and be modified by

them in return.

Affect is a useful lens through which to view nonplayer characters, as they are things that can

act, and be acted upon. Their bodies, such as they are, are changed during gameplay. The ways

in which bodies can change in games are determined by mechanics. So a study of how

nonplayer characters can act on the player, the game world, and each other, is a study of the

kinds of relationships game mechanics can afford. As the name implies, the position of the

‘nonplayer’ character is always tied to the position of the player.

In Chapter One we explore the limits of player and nonplayer actions and the affective

implications of these limits through the philosophies of Spinoza and Plumwood. Chapter Two

examines game enemies and their unique relationship to the player. In Chapter Three, friendly

nonplayer characters are interrogated about why they are so hard to design. The final chapter

1

looks at the representation of animals in games and how this representation of familiar beings

can influence human-NPC relationships. In concluding, this thesis explores a number of queer,

experimental, and indie games that use strange mechanics and uncommon roles to

demonstrate more interesting relational possibilities.

This study of affect between players and nonplayer characters in games necessarily

encompasses the power dynamics of player-nonplayer relationships. To fully appreciate the

nonplayer character, then, involves decentring the player and viewing the single-player game

world more as an ecology of experiences than a singularly human fantasy.

2

INTRODUCTION

Nonplayer characters of computer games feel things. Not the same things humans feel, and not

through the same mechanisms, but taking these feelings seriously can take us to some

interesting places. It is the purpose of this thesis to show that displacing players from the

centre of the game experience can lead to weirder and cooler games that operate as

ecosystems of affect rather than hierarchical arenas. Seeing players as just one part of a

complex whole can open up new ways of relating to digital others that isn’t wholly based on

the domination of one over the other.

This thesis comes from a discontentment with nonplayer characters (NPCs) that isn’t based on

the way they look or the way they move, but rather on the ways players are asked to judge and

act towards them. It posits that more interesting relationships between humans and NPCs

could be possible if humans weren’t at the centre of everything all the time. It tries to work

towards this decentering itself by examining the affective capacities of NPCs – the ways they

change and are changed. It uses affect and ecofeminist theories to explore how subjects are

formed in games through the power dynamics of humans and game characters. Overall, this

thesis hopes to be one possible answer to the vague question ‘what if we took NPCs seriously

on their own terms’ in an effort to achieve more interesting relationships with them.

During the long course of this thesis, the term NPC fell in and out of fashion as an insult

between real human people. The diss implies that one's opponent is robotic and programmed,

can only repeat rote phrases, and has no independent mind of their own. The insult relies on

the conception of the other as an instantiaton of a class rather than an individual, and is

basically another word for ‘sheeple’. The inferences and implications of this phrasing in terms

of inter-human connection has been discussed by Dafaure, among others, but the idea that

NPCs are repetitive, soulless automatons has not been so thoroughly addressed. That such a

sentiment can be so readily accepted points to it having some truth to it, but the easy

universality of such a statement suggests that it actually might be a self-fulfilling prophecy. If

3

players and designers understand NPCs as inadequate humans, less effort will be put into

creating and perceiving them differently.

Interactive, incorporeal, non-human beings are only becoming more common. Nonplayer

characters in games, helpful word-processing mascots, vtubers, virtual avatars of real devices,

virtual customer assistants, bots and chatbots, character accounts on social media, virtual

pets, virtual farm stock, shimejis and the like are already familiar to regular users of digital

media. The new relationships we have with these charming bodiless beings have not come out

of nowhere. Instead, the relationships humans have with the digital non-human are amalgams

of existing implicit knowledges and power dynamics, and have been adapted for virtual

spaces. In this way, relationships between humans and digital characters are connected to

existing ways of relating, but they are also something new and strange. While there is certainly

a link between the way humans treat their fellow reality beings and the way humans treat

digital beings, it is not a neat translation. Instead, the unique power dynamics of virtual spaces

create uncannily familiar ways of relating that hold up warped mirrors of our worlds to

ourselves.

The relationships that designers choose to encourage also say interesting things about what

they think about the implied or ideal users. The actions they program into their digital

characters reveal how they want the user to act. But the relationship the individual and the

character develop is dependent on a huge number of factors, including a character’s

representational shape, function, role and persona, the roles humans inhabit within and

without digital spaces, the expectations a human has for a character and their interaction with

it, and the possible actions the characters and humans can perform, all contribute to the

relationship between the character and the human. The relationship, in a way, has already

begun before the beings even meet. As humans increasingly come to interact and relate to

beings and objects that are designed with human use in mind, it becomes important to

examine the expectations we have for nonhuman beings.

Affect theory is an alternative to well-trod cognitive and psychological approaches to game

design that suggests new directions for the design of NPCs. The flow of affect between humans

and computer game nonplayer characters, follows a meandering path, and the relationship

4

between them is a node in a network of other turbulent connections. Players invariably have

reactions, interactions and feelings about nonplayer characters (NPCs) in games, even if the

reaction is dismissive, the interaction is brusque and the feeling is boredom. Humans have

relationships with digital characters, even if they are fleeting, subtle, shallow, or of a different

type or order than the relationships humans have to the beings that digital characters

represent. Relationships with NPCs are inevitable. The affect that mechanics afford may not

exactly be what the game designer intends either: players may be more panicked about

finding a save-point than any spooky monsters (Shinkle 2), and may feel more joyful about

getting to quit and have dinner than completing the final boss. Affect is not something that can

be poured into an empty game, but an existing quality that can only be managed, intensified,

or redirected. This makes designing affective nonplayer characters a bit of a challenge, and it

also makes the study of human-NPC dynamics particularly interesting. NPCs affect people

because everything affects everything, but there are particularities in this interaction that

have not been explored thoroughly yet. NPCs are made to be affected by players, and while the

outcomes of this affect are intended to be recognisable to humans, the mechanisms through

which affect happens are strange and nonhuman.

When games researchers talk about NPCs and affect, they talk about how NPCs aren’t doing it

right yet. To many games researchers, NPCs are a collection of problems to be solved. In the

literature they are seen as inadequate in two main ways: physical awkwardness (Afonso and

Prada; Arrabales et al.; Bailey and Katchabaw; Botea et al.; Boulic et al.; Cechanowicz et al.;

Creed and Beale; Huang and Terzopoulos) and bad representation (Beasley and Standley;

Curtis; Dickerman et al.; Downs and Smith; Fox and Bailenson; Hoffswell; Ivory; Loikkanen;

Malkowski and Russworm; Martins et al.; Miller and Summers; Pulos; Shaw; SundeAn and

Sveningsson). At the core of both of these criticisms is believability. NPCs are unrealistic, both

in their association of certain types of bodies with certain kinds of behaviour and action, and

their way of moving through their world. Women don’t all have huge breasts and wouldn’t

keep trying to walk through a wall, so an NPC that represents a woman shouldn’t look or act

that way either. However, the amount of literature detailing the inadequacies of NPCs shows

that even if NPCs are disappointing, they are imbued with a sense of possibility. They are not

irredeemably broken, but everyone who writes about them wants them to be better.

5

Though many games researchers fault NPCs for not being believable enough, this fault is not

entirely located in the NPC itself, but is a product of its position and action in its mechanical

and cultural environment. That humans use a mixture of social and programming codes to

interact with NPCs makes NPCs an exciting place where rules and assumptions can stack or

conflict. This makes NPCs fertile ground for a designer interested in affect. Spinoza offers a

way to tease out this tangled web of affection through his monist view of reality as comprising

of relations. His explanation of how affect is transformed into emotion through conscious

experience gives us an insight into how affective relationships between humans and NPCs are

experienced by conscious humans and unconscious nonplayer characters. The ecofeminist

Plumwood also helps to illuminate the minutiae of human/NPC relationships by defining one

of its most fundamental aspects: power. While her work centres nonhuman others like rocks,

plants, and animals, her explanation of how nonhuman otherness is created and maintained

can also apply to the design and interaction mechanics of NPCs. While computer game

designers may not have an overt intent to make computer models for ethical dilemmas, that is

often the eventual result (Schulzke 255).

From a Spinozan perspective, it doesn’t make sense to define NPCs by anything other than

their affect, as everything is “as much outside itself as in itself – webbed in its relations until

ultimately such firm distinctions cease to matter.” (Gregg and Seigworth 3). His philosophy

emphasises action – specifically, action that empowers or depresses. For Spinoza, emotion is a

result of an unconscious modification of the body that increases or decreases the body’s

relative power to act. And in games, any capacity for action relies on the rules of the game

world, or its mechanics. The mechanics of a game then, define not only a range of actions, but

also an emotional landscape.

The affective affordances of game mechanics for human-NPC relationships have not been

studied to the same extent as better representation and sophisticated movement. In games,

actions are tightly circumscribed, and everything in a game acquires an ‘effect tone’ (von

UexkuH ll 95) – an idea of how it can be used. Typologies of NPCs that clump characters together

based on their actions (Bartle 351; Isbister 229; Warpefelt ‘2016’) do so based on an

understanding that mechanics are the key underlying qualities that divide and unite.

Computer games don’t give players the ability to perform any action they want. Instead,

6

players have a limited repertoire of acts appropriate to the game genre. In this way, games are

like a workplace, art gallery, party, or other place in which context defines appropriate and

aberrant behaviour – just in a very concrete way. Like the player’s human body, the player’s

affective capacity is limited by the form it takes and the world that it’s in. The nature of a game

character is that there are always only a few ways to act.

von UexkuH ll’s work on umwelt, or how beings see the world and its affordances, shows how

the actions beings take affects their perceptions of objects. The actions a player can take in a

game affects how they see the game environment, and the characters in it. As Shinkle notes,

“Affect is key to the perception of images, and to the notion of meaningful interaction with

them” (2). When all the player has is an alchemist lizard mage with a glass mace, everything

looks like a hindrance to the quest. Further emphasising the role that emotion has in the

process of recognition, Spinoza notes that desire precipitates judgement (137). Knowing that

only certain kinds of interaction are possible with NPCs changes what players see as their

affective affordances, as well as their mechanical ones.

Flanagan argues that creating meaningful situations through mechanics is much more

effective than stating it in narrative (Flanagan ‘2015’). Persuasive ideas fed to players through

dialogue and in-game text is likely to appear heavy-handed and have the opposite effect than

intended (2). Ideas presented through mechanics are accepted much more readily. Sneaking

slowly through a room of guards induces anxiety, beating a boss is a thrill, and repeatedly

failing a tricky jump is frustrating. Mechanics is seen as the serious “core” of the game, the

authentic centre which is decorated by the fripperies of art and narrative. Mechanics is often

described as the soul of a game, and it is the mechanics which classifies a game as one genre

and not another. The high status of mechanics in discussion of games means that mechanical

solutions to problems are often seen as better than narrative or art solutions, by both designer

and player. Player actions are closely tied to genre, as their names make clear – first person

shooters, walking simulators, fighting games and racing games all refer to player actions.

Mechanics shape interactions between humans and NPCs, and so have a formative role in their

relationship. The player knows that the NPC is designed by a human and controlled by

computer programming. It cannot be separated from its narrative or ludic surroundings, and

7

instead functions as a temporarily embodied instance of a game-world itself. Yet the

conventions of story and sociability individuate the NPC somewhat individuated as an

affective agent in itself. It does not take much for an object to be perceived as having some

aliveness and personality (Reeves and Nass; Nass and Moon). In interactions with NPCs, the

player is always walking a line between knowing a computer’s artificial intelligence is behind

the character’s behaviour, but being called upon to act as if the character has real interests and

needs. For some players this is easy (Tavinor), but for others it brings up questions of how

best to treat others (Briggle), even when those others are fictional. The affordances of game

mechanics, then, have an important part to play in how the player can interact with an NPC

and what kind of relationship can be forged as a result. Interaction mechanics make the

difference between treating NPCs as embodied means to an end, or beings to be recognised.

The ways in which a human can affect an NPC has some bearing on how they feel towards

them.

However, what humans can do to NPCs and what NPCs can do to humans reveals a difference

of power. Players are made powerful in games by mechanisms similar to those that stratify

humans hierarchically in other contexts (Plumwood 3), while NPCs are defined by exclusions,

or what they are lacking. This difference is not harmful to humans or NPCs but indicates a

missed opportunity in the design of human-NPC relations. Dualistic conceptions of the player

and NPC (the player is real and the NPC is not; the player is free and autonomous and the NPC

is limited and glitchy; the NPC is created in ultimate service to the player; the player has an

important quest and the NPC does nothing important; the player can outrun and overcome

even the toughest NPCs with enough practice; the player has access to a larger range of action

than the NPC) are central to the problem of unsatisfying human-NPC relationships. Dualisms

aren’t always the best way to analyse everything (Haraway 6), but they work particularly well

for the simple worlds of computer games, because game design tends to create exaggerated

difference (Isbister 10). Power fantasies are fun, especially for people who aren’t ordinarily

powerful (Hammer and Baker). But power fantasies put players in a specific position and

orientation in relation to NPCs, and this allows and occludes certain affective possibilities. The

power imbalance between the implied player and NPC means that some relationships are

already closed off or directed into specific well-worn channels before the player even starts

the game. Disempowering the player, and repositioning them as part of an ecology of game

8

elements instead of at the centre, offers a way to make NPC interactions more affective. The

centred NPC can offer a new participation of the other in the self.

Most writing on NPCs to date talks about them in terms of types. NPCs are made to be useful,

and this usefulness is class-based. NPCs have key roles to play in setting mood (Greer),

indicating genre (Warpefelt ‘2015’), engaging players (Mallon and Lynch) and establishing the

game world (G. Smith). NPCs in games can vary from devoted companions with broad

dialogue trees to fleeting background decorations. But generally NPCs occur not as individuals

but as species, differentiated by appearance and behaviour (Warpefelt ‘2015’; Warpefelt

‘2016’). These types are defined by the relationship they have to the player. They might be an

enemy, an ally, or a resource, or subtype of these. The types are quite rigid, and their roles

quite fixed. While this categorisation results in one-dimensional characters and shallow

interactions, it also allows unambiguous judgement and quick action by players. The existence

of strong characterisation tropes also allows for interesting subversion of those tropes.

One of the oldest and most common types of NPC is the enemy. The enemy NPC has an

antagonistic relationship to the player-character and does its best to hurt them. The player in

return attempts to affect enemies to death. Enemies are ubiquitous in games because fights

are quantifiable, thrilling, simple to learn and easy to program. The player-enemy relationship

is characteristic of many computer game relationships in that it has a win state, interaction is

smooth and straightforward, progress is measurable, and success relies on quick impressions

and a familiarity with roles.

Although enemy NPCs oppose the player, it is friend NPCs that best illustrate the power

dynamics of human-NPC interaction. Friend NPCs help the player, give them resources and

things to do and are designed to be pleasant. They are also one of the hardest types of

characters to get right. Players interact with friends in very particular and sometimes oblique

ways, ways that aren’t especially enjoyable on their own. This means that friend NPCs have to

rely on other things to affect the player. Their affect cycle is intensified by time and progress.

Animal NPCs epitomise a few important characteristics of NPCs. They are decorative, they are

functional, they are useful, and they are often used as cover to avoid issues of human

9

representation (Isbister 119). They bring to light issues around what gets included in

reproductions of reality and how real-world power dynamics translate to the computer game

worlds. As nonhumans acting as another sort of nonhumans, they have a lot to say about how

humanness is constructed.

However, experiments into interesting NPCs, weird relationship mechanics and decentered

players have been attempted, and have been successful. Team Ico games (Ico, Shadow of the

Colossus and The Last Guardian) are a well-known example of affective NPCs, but queer and

indie games offer further insight into the possibilities of weird characters. The frenemies of

Moaning Columns of Longing show how to make affective antagonistic characters who

acknowledge their precarious dependency on the player. The companion of The Last Guardian

provides conditional support for the player while asserting his own boundaries, and the

absurd dialogue of Oblivion characters offers a learning opportunity about the unique charms

of NPCs. Rain World’s ecology of characters, in which the player is both predator and prey,

demonstrates a way to decentre players without disempowering player-characters. The

character of Ball refuses to recognise the player as an affective agent at all.

Early writing about nonplayer characters touted their affective abilities as empathy vehicles or

drivers of aggression. Both positions suppose that the game player is affected so much by

events in the game that they then start to engage in the same affective loop persistently, in

other situations, with other actors. The specificity of the relationship between the human and

the character in the virtual world is passed over in favour of relationships between humans in

the real world. In any case, the affective connection of player to avatar to representation of

others to real-life others has not been fully teased out, and the mechanism by which this

supposedly causal relationship operates has not been established. This thesis does not

comment on how relationships between humans and digital characters may influence

relationships between humans, but rather looks at the details of the human-digital character

interactions as interesting and worthwhile relationships in themselves.

When games researchers introduce their work they usually justify it by noting how much

money is spent on games (Afonso and Prada; Briggle; Dill and Dill; Fogel et al.; Fox and

Bailenson; Friedberg; Isbister), how much time is spent on games (Bulu; Chesney and Lawson;

10

Johansson), or that people feel things about them (Blom; Campano and Sabouret;

Coeckelbergh; Fox et al.; Garau et al.; Greer). What these metrics show is that computer games

are not apart from life but are a part of life, and as such they affect wallets, watches, and

people. This thesis addresses the kinds of affect and affective loops that can be generated by

and through humans and NPCs, especially those afforded by mechanics. It will focus on

English-language or English-localised single-player games where the player controls a player-

character. Multi-player games in which players interact with other humans as well as NPCs

enable an affective experience quite different to single-player games in which the player only

interacts with NPCs, and so while a study on the affective affordances of mixed human and

robotic otherness would be fascinating, it is outside the scope of this thesis. Similarly, games in

which the player is god-like puts the player in a very particular position in relation to NPCs. If

the possibility of mutuality with NPCs exists, the best place to find its beginnings is in games

where the player moves in a world like an NPC does, in a body similar to theirs. For budgetary

reasons, this thesis will also mainly discuss examples from free games, games that fortuitously

went on sale, games the researcher already owned, or games that have already been

thoroughly discussed in the literature.

11

CHAPTER ONE: SPECIAL AFFECTS IN GAMES

INTRODUCTION

In this chapter we shall take a tour of affect theory and have a look at some of the things it

might offer to an analysis of human-character interaction, starting with Spinoza and moving

on to more contemporary theorists. Through a review of current literature and quite a lot of

flowery metaphor, this chapter will try to reach a basic understanding of what a game

character is, what it does, how it does it, and even a little bit of why. Metaphors used in this

chapter include game characters as sense organs, as viruses, as coherences, as chimeras, as

robots, as puppets, as ghosts, as nonhuman animals and as conglomerations of cultural

references.

Everyone agrees that non-player characters are affective (Cole; Holtgraves et al.; Isbister &

Schafer; Korn et al.; Lankoski & Bjork; Lankoski et al.; Nass & Moon; Scholl & Tremoulet;

SchroH ter & Thon; von der PuH tten et al.; Warpefelt). What they disagree on is the ways in which

they are affective, whether players should be affected by non-player characters more, less, or

the same as they are now, and whether this affection has positive or negative, significant or

insignificant consequences for the way humans act toward other beings in reality. The

affective capacities of NPCs, like NPCs themselves, sometimes move smoothly, sometimes

meander into dead ends, and are yet to be fully appreciated.

Where great claims are made about computer games’ effects on humans, whether positive or

negative, they usually centre on the relationships that players and player-characters have with

their digital counterparts. Some, such as McGonigal, argue that games can increase empathy

and teach players how to see through the eyes of others. Others like Briggle argue that games

encourage violent behaviour and lead to objectification. In discussions of games’ broader

effect on human psychology, relationships with non-player characters are central. But the

mechanisms for these purported effects has mostly been discussed in psychological or

cognitive terms. And where the effects humans have on game characters are discussed, they

12

are talked about in terms of game design and programming. With its focus on relationships

and interactions rather than individual humans, the nascent field of affect theory can provide

a unique perspective into the strange affective ebbs and flows of humans and game characters.

The majority of research in the area of relationships with human-made non-human

interactable thingies has been concerned with two things: whether violent interactions in

computer games cause violent interactions in real life (Anderson; Anderson & Bushman; Dill &

Dill; Ferguson; Gentile et al.; Sherry), and how to make robots more likeable (Bates; Ben

Mimoun et al.; Coghlan et al.; Coeckelbergh 2014; Duffy & Zawieska; Hall; Holtgraves; Jones &

Deeming; Lim et al.; Nass; Riek & Howard). These have limited but important insights to the

subject of this thesis. Media theories also has some contributions to make in regards to

representation and the perceived and actual realness of in-game actions. When game

designers get involved in this topic, research tends to be concerned with how to make

nonplayer characters more intensely affective (more threatening, more cute, more friendly,

more hateful etc) (Isbister), taxonomic surveys of the kinds of services nonplayer characters

provide to humans (Johansson; Warpenfelt and Verhagen), what kinds of people get

represented in games, and how stereotypical they are (Beasley & Standley; Burgess et al.;

Downs & Smith; Ivory; Martins et al. 2009; Martins et al. 2011; Miller & Summers), how

nonplayer characters can be less annoying in terms of dialogue (Morrison & Martens),

doorways (Huang & Terzopoulos), walking around (Botea et al), and basic awareness (Kasap

et al), and instead more helpful (Guckelsberger et al), or if digital characters can consistently

provoke specific emotions in interactions with people at all (Isbister & Schafer). That is, game

designers have mostly engaged with the question of affective interactions with nonplayer

characters by examining specific mechanisms through which affection could possibly be

improved.

AFFECTION FOR SPINOZA

Spinoza’s definition of affect asserts that every physical thing, living being, idea, mental state,

(and anything that can possibly be conceived of as an object really), can, and does, modify each

13

other. The ability to change or be changed does not just reside uniquely in humans, or even

living beings, but is a quality possessed by just about everything imaginable. A human can

affect another human, but an oven can affect cake batter, rain can affect a dandelion, and a

wombat can affect rocks. Modification by other things is an inherent part of being a thing, and

relationships are part of existing, Spinoza asserts. This is how Spinoza’s particular brand of

affect theory becomes useful for this thesis. His monist flattening of subject and environment

cuts through the human-centrism of previous work done on human-NPC relationships and

offers a radical new perspective where humans and NPCs can be viewed in terms of their

effects rather than their relative moral status. And through modifying others, things

themselves are modified. It follows then, that digital characters can affect humans, and

humans can affect digital characters. And this affect, once processed through the human body,

can result in the conscious experience of emotions. But this affect, once processed through a

computer, can also result in the setting and changing of predetermined variables that instigate

particular character behaviour. So although the digital character does not have the conscious

experience of emotion, it is still affected by the human player, just as it can be affected by its

environment, or any stimulus it is programmed to be sensitive to.

Many players find the ability to affect and be affected pleasing. And pleasure, according to

Spinoza, is an emotion arising from the fact that the body has been modified to become more

powerful (138). Pain is the opposite, an emotion arising from the body becoming less

powerful. For Spinoza, what is pleasurable is what is good, and what is good is what is useful,

and what is useful is what brings the body to a greater state of perfection, and what is

perfection is being able to decisively act in line with one’s nature: “What I mean [by greater

perfection, is] that we conceive the thing’s power of action, in so far as this is understood by

its nature, to be increased or diminished” (190). Bodies always want to feel pleasure. The

experience of pleasure means that the experiencer gained a greater ability to act in the world,

to act according to their true desires. “Emotion,” Spinoza argues, “is an idea, whereby the mind

affirms of its body a greater or less force of existence than before. When, therefore, the mind is

assailed by any emotion, the body is at the same time affected with a modification whereby its

power of activity is increased or diminished” (195). Something that increases one’s ability to

act can be said to increase one’s power, while something that decreases one’s ability to act can

be said to decrease one’s power. Power, ability, and affect, then, are inextricably linked, and as

14

games are primarily power fantasies (Hammer and Baker; Parkin) they are an extremely

interesting place to put affect theory to use.

Spinoza maintains that there are only three basic feelings: pleasure, pain, and desire. These

correspond to relative increases and decreases in power. All other emotions, he argues, are

just names for derivatives that arise in particular circumstances. An individual’s power to act,

then, is always tied to some emotion, as “all emotions are attributable to pleasure, to pain, or

to desire” (227) and these feelings always instigate changes in one’s ability to act: “Pain

diminishes or constrains man’s power of activity, in other words, diminishes or constrains the

effort, wherewith he endeavours to persist in his own being … pleasure increases or aids a

man’s power of activity” (155).

It can be tempting to dream of a general theory of affect with affect fields and subatomic

affectons and laws of emotion and the possibility of a giant atmospheric bomb. The

rationalistic geek culture surrounding games can invite a researcher to appeal to hard and

cool neuro- or cognitive sciences for support in their conclusions, or at least to try and sound

sciencey (as Plumwood notes, hard, mechanistic, masculine types of knowledge are more

likely to be taken seriously (117)). But the key strength of affect theory is that despite

Spinoza’s admiration of Descartes, he rejects the latter’s mind/body dualism and expertly

vaults over previous walls that have been built between physical and mental phenomena.

Affect theory is concerned with relations, and the particular objects that are in relation matter

less than the nature of the connection. It is seductive to describe affect only in terms of its final

outcomes on some machine or person, rather than the transformative process itself. That’s not

to say that a study of the outcomes of affect is irrelevant. Just that a focus on the objects and

subjects of affect brings the focus back to individual entities, when it should be about the

betweenness. That said, this thesis is not about the nitty-gritty of machine code or neuron

firing. As Nitsche notes, “the cinematic and visual rules on the mediated plane affect the

player’s comprehension of the game space … code alone is not sufficient to describe and

define game spaces and the experiences they provide” (28); game design utilises moral and

social codes as well as programming codes.

15

Affect in more recent times has been discussed as something that is pre-individual, from

which emotions are only one outcome of many (Massumi). Nash has described affect as a

turbulent ocean of forces through which individuals are always about to come into being but

never settle, while Gregg and Seigworth see it as “the name we give to those forces … that can

serve to drive us toward movement, toward thought and extension, that can likewise suspend

us (as if in neutral) across a barely registering accretion of force-relations, or that can even

leave us overwhelmed by the world's apparent intractability. Indeed, affect is persistent proof

of a body's never less than ongoing immersion in and among the world's obstinacies and

rhythms, its refusals as much as its invitations.” (1). That is, affect is in-betweenness, and so is

everywhere and everything. It’s not necessarily good, and refers to the process of change as

well as the change itself. While it is colloquially analogous to ‘emotion’, in this thesis the term

‘affect’ is used more precisely as any change in the body, which may or may not register

afterwards as a conscious emotion. Affect is always occurring, because bodies are always

changing in relation to their environment. But this affect can change in direction or intensity,

depending on the objects and relationships in play.

THE POWER OF SPINOZA

“For Spinoza,” Armstrong writes, “the mark of more complex and powerful individuals is a

heightened capacity to be acted on and affected by other bodies, and that the mind’s power to

think and, therefore, to form adequate ideas, increases in proportion to its body’s capacity to

be acted on” (57). Gatens further observes:

“how it is we become conscious of our power to affect and be affected in ways that

cause joy or sadness, that is, in ways that involve an increase or decrease in our feeling

of power. The more our actions emanate from within us rather than outside us then the

more powerful and free we are and the more we experience joy. The capacity to act,

rather than be acted on, to express one’s own nature rather than merely reflect or react

to the nature of another, is an expression of one’s freedom, power, virtue or conatus.

16

Throughout the progress of the Ethics, these terms come to stand in an almost

synonymous relation to each other” (Gatens 199)

Like humans, what is properly inside and outside an NPC is debatable. But determining

whether an NPC is self-motivated is a little easier than it is for humans, because (if the game is

open-source or if the code has been released), the code for the NPC and the game can be

analysed to see how sensing and acting works, and where it occurs. So whether actions

emanate from within an NPC’s ‘self’ or from outside can be determined by looking at its code

(though as Nash notes, the protocols by which a digital entity acts are always determined by a

human in the end (4)). But an NPC does not feel joy, the experience of power or freedom,

because it doesn’t care what makes it act, whether it acts at all, or whether it is acting more

effectively. However, as Gatens notes, the feeling of power may not always relate to whether

one actually does have power:

“Many of our affects give rise to feelings of power that may be largely imaginary. Yet

these imaginary feelings of an increase in our power may indeed increase our power to

act, though this is likely to be in an erratic and unreliable way. Such joyful affects may

easily be reversed and hence become sad and debilitating affects.” (Gatens 199)

Players have less power in a game than they do in other contexts, since they can only perform

very few in-game actions. They can purchase a sword, run outside, and kill a wolf, but they

can’t go to the post office, make friends with a wolf, or have an interesting conversation with

anyone. The power they do have is the power to act without consequence. The few in-game

abilities they can perform intensify quickly and noticeably, so the feeling of gaining power is

strong, even if the player is doing the same thing as they were an hour ago.

‘Power’ is a tricky word though, because it refers to so many types of difference. Spinozan

power is a quality of an individual (insofar as there are any individuals in this swirly world of

constant transformation (Armstrong 60)) rather than something that someone has over

someone else. But games trade heavily in power as both individual and relational qualities. A

player has the power to shoot a fireball at a policeman both in the sense that they are able to

17

do it and also in the sense they are able to get away with it. Power-to and power-over are not

so different in game worlds.

Digital characters are powerful in a number of ways. They are creatures of electricity, many

can perform superhuman feats, and they can compel other beings – as we will see in a

discussion of Horizon Zero Dawn’s hierarchical system of oversouls. But what Spinoza means

when he talks about power is the strengthening of the ability to act, and particularly to act in

ways that express one’s true self. As Armstrong observes, a Spinozan view of the self is more

about alignment and the maintenance of clear, strong flows of affect than of clear boundaries

between self and other: “If the Spinozist individual is her ongoing bodily and mental relations

with the surrounding environment, it follows that not all otherness can be regarded as

external to her. Spinoza’s relational conception of individuality involves a fundamental

rethinking of concepts of interiority and exteriority that are premised on limiting the borders

of the self to the spatial boundaries of the anatomical body or the self-enclosed confines of

Cartesian consciousness” (56). The digital character, which has a body (as it were) distributed

between various subroutines, visual display, cultural stereotypes, genre expectations,

electricity and pixels, poses no problem as a subject of affective analysis.

A digital character, then, does not have to be conscious or human-like to have and wield power.

It only has to have some essence to preserve, some preferences to align to. Armstrong notes

that Spinozan otherness is more akin to friendship and mutuality than physical proximity:

Interiority and exteriority are instead redefined in terms of what “agrees” or

“disagrees” with the individual’s identity. What is internal to the individual is that

which agrees with her nature in the sense that it contributes to her self-maintenance or

augments her power. On the other hand, what is external to the individual is that whose

nature is contrary to her own in the sense that it opposes her striving to persevere in

her being (E, III, Props 4 and 5). Thus, my relations to others who “agree” in nature

with me are internal to me, whereas the virus in my body, although it is spatially inside

me, must be regarded as an external cause.” (56)

18

Affect in digital spaces then, has to do with the power to perform tasks that align with one’s

character. But because the player’s character is at the centre of events, this usually results in

power over others. When games are referred to as power fantasies, the power that is

fantasised is both physical power (to jump high, to run tirelessly, to fly, to punch real hard) and

relative power (to order around, to disregard, to train, to punch real hard). It is inevitable that

in achieving one sort of power, the other will follow.

Game designers appear to have taken Spinoza’s idea of power completely literally, and so

sandwiches, swords, skills and social interactions affect the player’s power instantaneously,

linearly, and measurably. Many single-player games start the player in an outsider position of

relative powerlessness (a prisoner, an outcast, a visitor, an amnesiac; level 1) before the player

is set free to become the mostest bestest person ever in the game world. This transition of

nobody to VIP parallels the player’s mechanical experience (of clumsy controller fumbler to

combo-move master), and exaggerates the ubiquitous story of rags to riches that many games

rely on. However, the player already knows they are better than digital characters, because

they have organs and free will and things, so the initial situation of being low in a game’s social

hierarchy can never be very intense. What can be felt, though, is the inevitable demonstration

to the inadequate inhabitants of these petty worlds that the player was better than them all

along. The function of the power fantasy is not to make the player feel more powerful, but to

confirm their existent power. Power, then, is revealed as contextual.

Games are commonly referred to as power fantasies (Hammer & Baker; Parkin; Pellechi). They

are admired for their ability to make players feel competent and accomplished in a world

where many feel the opposite. But a game world is a place where a human can barely perform

any actions at all, nothing of importance is affected, and any actual changes made are

reversible with the right save file. Schell notes that in games, a player’s “verbs” are limited,

that is, “Games usually limit players to a very narrow range of potential actions, while in [book

and film] stories the number of possible actions that characters can engage in seems nearly

limitless” (143). Yet the act of temporarily inhabiting the body and role of a being who is

relatively powerful to their context is compelling. Player characters can do very gymnastic,

destructive or creative actions beyond the abilities of ordinary people in non-game contexts,

which is super cool. But it’s quality over quantity; they can’t perform the thousands of acts

19

that an ordinary human can. The power fantasy certainty isn’t the fantasy of being able to

perform a wide range of motions or act outside of genre. It is the fantasy of differential. Player

characters have limited abilities but nonplayer characters have even fewer. Players are able to

affect the game world, and to be affected only in ways they expect or consent to. By definition,

any (designed) obstacle encountered can be overcome and any enemy bested (even if it

involves some tedious grinding). A game constricts a player by giving them limited abilities

and usually an intended narrative, but once entering that pond, a player is the biggest fish. The

power fantasy isn’t to increase power, but to have more power relative to others. And the

easiest way to do this is to make a game in which the abilities of the nonplayer characters do

not match that of the player. A player goes into a game wanting to be affected, and wanting to

have a meaningful effect on others. Game characters could easily be unbeatable (as Eichner

points out, this is standard for child characters) but this would not serve the affective aims of

the designer or player. A game like Ball, in which the player can have no meaningful effect at

all, is boring to some and infuriating to others. A game like LIM, where enemies hold all the

cards, still gives the player unique abilities. The 1962 nim-based game Marienbad in which the

player plays against a computer opponent was actually unwinnable and apparently both

frustrating and addictive to play.

WHAT IS A DIGITAL CHARACTER?

Strict definitions aren’t too helpful here, but generally, NPCs are characters in a computer

game whose actions are not directly controlled directly by a human, but determined by

computer programming. Digital characters are chimeras of programming code, invisible

skeletons, bitmap textures, animation operations, mental associations, narrative conventions,

variable stories and stored variables. They can be seen, heard, manipulated, conversed with,

petted, helped, fought, killed and eviscerated – but not smelled, tasted or touched (toys-to-life

games notwithstanding). They cannot suffer, die, rejoice, or be free, though they can

decompose and transform into something else. They have recognisable faces but are hollow

inside, having no inner world, no subjective experience and no autonomy or will of their own.

They have no life or interests, but players are called to act as if they did. Players interact with

20

digital characters in familiar ways, using everyday social tools: talking, exchanging objects,

gentle or violent actions. Digital characters may be affective because they look cute or

threatening, but they are also affecting because players are called to act towards them in

caring or aggressive ways. They are coherent beings in that they cohere out of multiple things

and this conglomeration is recognisable within subject-object frameworks as a distinct

individual. While game characters cannot be said to be conscious (or, in Spinozan phrasing,

operate with adequate causes (129)), and they don’t have emotions (modifications of the body

of which they are aware (130)), this does not matter in terms of defining them as beings, as in

a Spinozan view, “bodies are distinguished from each other in respect of motion and rest,

quickness and slowness, and not in respect of substance” (93).

From the point of view of a designer, it can be convenient to view nonplayer characters as

beings as individual as the player, both for the benefit of code and for the benefit of making

realistic characters. Schell asks “whether virtual opponents, from a game mechanics

standpoint, should be thought of as players” (138), and suggests “from a private/public

attribute point of view, it makes sense to consider virtual opponents as individual entities on

par with players” (138) (‘public/private’ being a programming term referring to the location

that details are stored). While a digital object cannot “know” something, saying that Pacman's

ghosts know the location of Pacman can be a useful construct when designing pathfinding

ability. Even if players don’t think of digital characters as agents, it can be useful for designers

to conceptualise them that way, to design their behaviour.

Digital characters cannot think and are not conscious, a quality which is used to differentiate

between ‘the thinking being and mindless nature, and between the thinking substance and ‘its’

body, which becomes the division between consciousness and clockwork’ (Plumwood 116).

The common perception of digital characters as bodies without minds leads into well-trodden

territory. This territory is a Cartesian paradise. But just because characters do not think does

not mean they do not perceive and sense their world. The NPC is often dismissed as an empty

shell– a ‘body … [which] becomes an empty mechanism which has no agency or intentionality

within itself, but is driven from outside by the mind’ (115).

21

Seeing NPCs as hyper-separated Others subsequently allows them to be folded into the

player’s Self. Once characters are seen as mindless bodies wholly different to the player they

are “alien to the self, [which] excludes the possibility of mutual recognition or exchange’

(Plumwood 117) and can be exploited, that is, subsumed. This dualism is ‘a mode of attention

to the world which denies dependency and kinship between observer and observed’ “(123).

This practice is built into the design of game relationships. Digital characters exist to be

incorporated into the player’s self. The mindless character exists to be used, like a toaster, or

any other machine. Thus the townsperson becomes a source for a quest, the rabbit becomes a

source for rabbit meat, the angry wolf becomes a source for entertainment and a source for

wolf pelt, the ally a source of free labour. They become property to be owned, territories to be

annexed into the larger body of the player. In these subject-object relations it becomes difficult

to ‘grasp the aliveness of the other … [it becomes] a relationship of opposition and power’

(Plumwood 122). It is hard to imagine true dependency and kinship between beings as

different as a human and a fictional digital character. It is difficult to see how they could be

comparable at all. But this thesis is not arguing for game character liberation. Changing player

and designer mindsets towards digital characters opens up possibilities for new kinds of

relationships. There is no need for general equality between humans and digital characters,

but better relationships can develop when humans and characters are interdependent during

a time, or equal for a task.

Digital characters are almost always representations of other beings. Their appearance or

behaviour can be a cue for the player in learning how to interact with them. Therefore part of

what it is to be a digital character is to be a referent to other things. They are evocative. They

are a stand-in for something else. Like moths with eye-spotted wings, they are unknowing

carriers of meaning. And like the blackbird who is startled away from his lunch, the human

player automatically registers the friendliness or dangerousness of the digital character’s

presentation. This is a key part of their affective ability. Isbister notes that agreeableness and

dominance are two of the most basic judgements a human makes when they first meet

another being (25), and states that this first impression is hard to modify. The player is

affected by the ideas the character evoke, as well as the object in front of them. Spinoza notes

that what is evocative for one person may not be for a different person (163), and so digital

characters rely on specific shared cultural settings for their meaning. But characters also face

22

an opposite pressure to appear unique, as “an object which we have formerly seen in

conjunction with others, and which we do not conceive to have any property that is not

common to many, will not be regarded by us for so long, as an object which we conceive to

have some property peculiar to itself” (Spinoza 164). The appearance of the character then,

must find a balance between being familiarly intelligible and strikingly new. Spinoza’s first hot

design tip, then, is to make characters look more or less unusual depending on how much you

want them to affect the player with “veneration” (preoccupation with a uniquely lovely object)

or “consternation” (preoccupation with a uniquely hateful object) (163).

This is demonstrated nicely by Adam Nash’s Moaning Columns of Longing in the online virtual

environment Second Life. This interactive artwork shows that beings do not even have to be

visually representative of a living thing in order to be affective. Nash describes the work as

such:

“When a user’s avatar approaches the work, a virtual ‘column’ is spawned, glowing

white and spurting glittering particles of joy while it declares its undying, faithful love

for the user via the ‘chat’ facility built into the interface. Once the user leaves the space

(that is, logs out), the column starts sending more and more emotionally manipulative

emails begging the user to return. Every hour that the user does not return to ‘touch’

the column (‘touch’ is the Second Life interface analogy used instead of ‘mouse click’),

the column becomes a little duller, a little shorter, and starts emitting a moaning sound

that becomes louder as the hours pass. The hourly emails become more desperate and

more emotionally demanding (“without you, I will die”). If the user does not return

within 24 hours, the column will ‘die’, that is, in technical terms it will be deleted from

the database permanently. If the user does return to ‘touch’ the column within 24

hours, the column returns to its full height, glowing brightly and rejoicing loudly over

the chat channels, “[username] loves me! My existence has meaning!”. Over the couple

of years this work was in operation, some users maintained their columns for months

at a time.” (Nash 2012 24)

This column is a geometrical structure, makes no claims to be anything but digital, and its

non-existence has no deleterious effects for the player’s game. The columns certainly aren’t

believable as anything other than a digital object with chat capability. But the response of

23

Second Life players indicates that there are other factors than believability in the affective

capacities of digital characters.

Hui’s Aristotelian analysis of digital objects provides a contrast to affective analyses. He argues

that digital objects, including digital characters, are completely different from natural objects,

and are perceived and treated as such automatically. Natural objects are objects which are

perceived, comprehended and appreciated as existing in a world of unified meaning, and have

a ‘thingness’ about them – a sense that there is such a thing as being that thing. Although

digital objects appear to be natural objects, at a closer look they have no core. They “appear to

human users as colourful and visible beings, yet at the level of programming they are text files;

further down the operational system they are binary codes, and finally, at the level of circuit

boards, they are nothing but signals generated by voltage values and the operations of logic

gates … we may end up with the mediation of silicon and metal” (29). Hui argues that since a

game rabbit has no physical location, shape, colour, firm physical presence or fixed appearance

it is more akin to a set of instructions or an idea, than a real rabbit.

This perspective, however, does not detract from affective ways of thinking about digital

objects. While Plumwood would observe that a focus on definition creates the potential for

overdetermination and so stratification and hierarchisation (53), whether digital characters

have a solid thingness actually doesn’t matter in the question of whether they are affective

(though maybe Aristotelian players will only have platonic relationships). Although characters

might be an idea or a tool, unconscious and incorporeal, they can change, and they can be

made stronger or weaker. Even a virtual rock in the corner of the most unrealistic and obscure

game can be part of affective relations, if it is destroyed, if it provokes some idea in the player,

if a character navigates around it, if it is shown in pixels on a screen, if it is glitched by some

corruption of the game program, if an artist spent some time on it, if it becomes a meme.

AFFECTIVE STRUCTURES OF HUMAN-NONHUMAN RELATIONSHIPS

The promise of games is that of an affective exchange – that in return for affecting the game

world in certain ways the player will receive certain affective rewards. For this to work, both

24

the methods and rewards have to be unambiguous. Discourse around affective nonplayer

characters define the value and success of a character by how intensely it affects the player

with affects intended by the designer. There is pressure to make characters that always affect

players in predictable ways. Like getting characters to pick up and hold teacups daintily, this

often requires either a lot of complex programming, or a cinematic specificity of position and

inflexibility of movement. In his study of perception and meaning, the biologist von UexkuH ll

suggests that objects acquire ‘tones’ of potential action, specific to the species and individual

looking at them. A cat might see a laptop as an obstacle, or a warm sleeping pad, for example,

while a human might see it as entertainment or a workplace. If the cat has been taught that

they’ll get in trouble for sitting on the laptop, the tones of the laptop can change. ‘Tones’ of

potential relationships also exist when it comes to digital characters. Affective flows between

players and nonplayer characters are channelled in certain directions by player expectation,

genre convention, and a certain homogeneity in design. Many things other than the immediate

interaction colours the relationships of players and characters. As Leino notes, the overall

attitudes, mood and mindset of the player affect the potential affections between players and

characters. As well as mechanical affordances, games also have affective affordances in that

players may perceive that only certain relationships are possible with nonplayer characters.

“Consciousness is always directed toward objects and hence is always worldly, situated, and

embodied” (543) states Ahmed, indicating that humans are always attending to an object,

always oriented towards a thing, and so consequently, oriented away from other things.

Further, Ahmed argues that the value of objects is culturally determined, and that “being

directed toward some objects and not others involves a more general orientation toward the

world. The objects that we direct our attention toward reveal the direction we have taken in

life. If we face this way or that, then other things … are relegated to the background” (547).

Since humans cannot possibly perceive everything all at once, some objects will always be

more attended to and valued more than others. If games intend the player to attend to some

things more than others, or take a position in relation to others, this reveals what they want

the player to value and ignore.

Feelings cannot be quantified as precisely as a high score. But display of intense emotion is an

integral part of social gaming, game criticism and marketing, and new genres of serious or

25

meaningful games seek to expand the range of emotions that games can ‘produce’. The

utilisation of affective game design reaches its zenith in gamification. Gamification promises

that the affect ‘produced’ by games has long-lasting benefits for productivity in the workplace

and at home. This works towards emotions, Ahmed argues, being “re-presented as a form of

intelligence, as ‘tools’ that can be used by subjects in the project of life and career

enhancement” (3). Games as productivity tools promise to cultivate “good emotions”, which

“remain defined against uncultivated or unruly emotions, which frustrate the formation of the

competent self” (3). Some games then, promise to be so affecting that they can change the

player into something better. But as paolo of MolleIndustria observes in his critique of Games

for Change, this ignores games’ status as “objects we can think with” in favour of them being

seen as “mere vectors for messages to be dumped into players’ brain”.

Players are put into particular positions so they have certain orientations relative to

characters, and relationships have certain tones, even before they meet. This is inevitable to

some extent, but the particular positions, orientations and tones intended by designers often

mirrors those of non-game contexts.

HUMANS AND NONHUMANS

The Australian ecofeminist philosopher Val Plumwood posits that Western cultures are

infused with the mindset of hierarchical dominance, in which a few are the master and many

are the servants. This mindset places living beings in categories, arranges these categories in

hierarchies, and then draws associations between categories and conclusions about

individuals from these placements. Plumwood’s insight that non-humans can also be included

in this dynamic, such that animals can be judged as passive as females, as irrational as the

colonised, or as unproductive as children, illuminates a reason/nature or human/environment

binary. This leads to some interesting places when applied to constructed, human-designed,

but nature-referential, digital spaces.

Plumwood notes that there is a common pattern to domination, whether it is men’s

domination of women, the coloniser’s domination of the colonised, or humans’ domination of

26

nature (141). In all processes of domination, instrumentalising the dominated is a key step.

Defining instrumentalism, Plumwood writes:

“Instrumentalism … is the kind of use of an … other which treats it entirely as a means

to another’s ends, as one whose being creates no limits on use and which can be

entirely shaped to ends not its own … a mode of use which does not respect the other’s

independence or fullness of being, or acknowledge their agency. It recognises no

residue or autonomy in the instrumentalised other, and strives to deny or negate that

other as a limit on the self and a centre of resistance.” (142)

This process is also apparent in computer games. But in an interesting twist of fate, computer

games and the others within it are actual instruments. It is difficult to argue against

instrumentalising an instrument. There is no self to recognise, there is no agency to crush,

there is no will to bend. Researchers who have identified this pattern find themselves falling

into one of to two classic arguments: “isn’t it bad just to do these things even though no one’s

being hurt?” (Whitby) and “maybe just acting in this manner is harmful to the actor’s moral

character?” (Briggle). But there is a third objection to be made against games’ perpetuation of

the mastery mindset: it’s overdone and boring. When Plumwood observes that “the structures

of self involved in human domination and colonisation are reflected, repeated and confirmed”

(142), she refers to human relations with nonhumans like animals, mountains and glaciers.

But these structures are also “reflected, repeated and confirmed” in humans relations with

nonhumans like nonplayer characters. Unlike animals, mountains and glaciers, nonhuman

nonplayers are not harmed by this relation. But as a result of it, they are dull. Game characters

could be more interesting if they moved away from this familiar dynamic.

This positioning of players and nonplayer characters in a hierarchy influences the affective

flows possible in a game in two ways: the means through which players and nonplayer

characters can affect each other, and the specific affects that result. The steering of affect into

completely chartered waters has consequences for the construction of the player as a being in

the game and the subsequent relations they can have with nonplayer characters. One of the

consequences of this is the creation of the player as a being separate and superior to other

beings in the game. Quoted in Plumwood, Benjamin notes the construction of the independent

27

self requires the management of affect and specifically the management of affect flows: the

self “continues to move in the realm of subject and object, untransformed by the other. The

self says ‘You cannot affect or negate my identity, you can only be the object of my assertion’”

(145). Plumwood continues: “In instrumentalised relations to the other, the self takes no risks

but is not open to real interaction with the other, because the independence of its desires

makes the dualised individual a closed system. The definition of the other entirely in relation

to the self’s needs means that it is encountered only as incorporated by the self” (145). The

player-as-master results in nonplayer character being defined in terms of the services they

provide to the player, and player-characters as individuals isolated from mutual dependencies,

being positioned as outcasts, prisoners, orphans, chosen ones or otherwise outsiders. This not

only lets the player experience the exciting narrative “I used to be the least cool guy but now I

am the most coolest” but excuses them from any kinship, or even civil, obligations. The

nonplayer character becomes something that is affected more than it can affect, except as a

tool the player uses to affect themselves.

Dominance need not appear in the forms of exploitation, cruelty or hostility. It only requires

that the dominated are seen in opposition and irreconcilable to the dominant. Plumwood

suggests that there are a few methods used to maintain this situation. Backgrounding involves

“treating the other as background to the master’s foreground” (48), denying the importance of

the other and their work, or dismissing them as not worth attention. Hyperseparation is a way

of emphasising difference between two beings, magnifying any differences and minimising

continuity. It results in the idea of essential natures “which explain and justify widely differing

privileges and fates” (49). Incorporation makes complementary and opposing these essential

natures; what the master is, the slave is not, and vice versa (52). The lower in the hierarchy “is

construed not as occupying a space on her own account, but as enclosing a space for another”

(52). Instrumentalism is the most relevant for this thesis. Instrumentalised others “are

conceived of as his [a master’s] instruments, a means to his ends … [the master] does not

recognise them as moral kin, and does not recognise them as a centre of desires or needs on

their own account … the underside has no intrinsic value, is not for-itself but is merely useful,

a resource” (53). Homogenisation imagines the dominated class as all the same, all alike.

28

This all sounds very serious for a discussion of silly computer people. But even though

nonplayer characters don’t have feelings and can’t be hurt, they can still be part of the same

familiar logic that affects other kinds of relationships in other spaces. Nonplayer characters

are always defined by a lack – they are not the player character. Nonplayer characters are quite

different from humans and probably shouldn’t be given equal rights. Plumwood notes that

“dualism should not be seen as creating difference where none exists. Rather it tends to

capitalise on existing patterns of difference, rendering these in ways which ground hierarchy”

(55). And because there are differences between human players and nonhuman nonplayer

characters, this logic finds an easy place to settle in.

Plumwood asserts that objects should not have to be human-like in order to be respected as

agents (136). She argues strongly that nonhumans should be appreciated on their own terms

without having to demonstrate relatable human characteristics like consciousness, self-

awareness, speech, souls, or aliveness. Searching for such things to admit nonhumans into the

human moral community, she says, denies the essential otherness of the other. It means that

nonhumans can only be appreciated by how similar to humans they are, further entrenching

the primacy of the human. What we need, Plumwood suggests, is a way of conceiving others

that does not rely on the humanness of the nonhuman, but a recognition of kinship and

directed activity. Here’s where it gets tricky.

Plumwood does have criteria to allow certain kinds of things into her moral community. And

her distinction is whether the object can demonstrate an intended direction: an intentionality.

‘The Coca Cola bottle when individuated as part of a human instrumental context is not

(or not without a further context) framed intentionally, and we can neither hinder nor

assist its journey. Unless it is individuated as an artefact, that is, subsumed within a

context of human agency and intentionality (in which case it is subject to the

considerations of intrahuman ethics), there is no obvious intentional context to place it

in. Unless we can find one, it has no direction of travel, and all outcomes with respect to

it are indifferent. But the glaciated valley can easily be conceived as such an intentional

system, if considered as part of a directional, developmental process of the earth, and

we might both hinder its journey and stop it telling its story by damming it ’ (138).

29

However, this relies heavily on the imagination of the human observing the object. Coca Cola

advertisers would no doubt tell a wonderful story about the plucky little plastic bottle and its

deep desire to be purchased for money. What beings are deserving of respect can be a matter

of marketing. Games offer an imaginary, low-stakes spaces where humans encounter a variety

of nonhumans and experiment with different ways of relating to them.

According to Plumwood, a nonhuman should not be treated with respect because it will be

grateful, a spiritual force demands it, or that it is secretly a human underneath its

fur/bark/bitmap texture, but because acknowledging kinship with nonhumans is a more

interesting, accurate, and fulfilling way for humans to be in the world. Recognising nonhumans

as kin does not mean their needs have to become ours or our aims in life have to be the same

(214). ‘We can be delighted that our local bandicoot colony is thriving without ourselves

acquiring a taste for beetles’ (160) after all. But kinship demands questions: “We must

interact, but how far as I entitled to assert myself against or impose myself on the other? We

must adapt to each other, but is one party always to be the one who adapts the other other to

be adapted to? How much must we leave for the other? How much can we expect to share?

Here there is not just one play of exchange between self and other, but multiple contextual

ones” (139). These questions have one predetermined answer if one is the master and the

other is the dominated, but a multiplicity of answers is possible in a framework of respect.

And these answers might not be fluffy happy ones either. Haraway notes that “none of this

work is about finding sweet and nice - “feminine” - worlds and knowledges free of the ravages

and productivities of power. Rather, feminist inquiry is about understanding how things work,

who is in the action, what might be possible, and how worldly actors might somehow be

accountable to and love each other less violently” (7). This suggests that in order to create

digital nonhuman that are not just tools to achieve a prearranged outcome, these characters

must be created with independent needs which are contextual to the game world and not the

player. They must be “autonomous intentional systems” (Plumwood 138) and presented that

way to the player.

30

But the question of whether a designed object can be recognised and respected at all is still

under doubt. Briggle identifies limits on power as a necessary component of relating to others

well. He argues that when games position human players as powerful, they may be hindering

players’ ability to relate to others virtuously, or even recognise them as others at all (172).

Because respecting otherness requires one to place limits on oneself in recognition of

another’s needs, powerful game players are not recognising digital characters as others

because they do not have to limit themselves in interactions with them. Briggle argues that

losing control, facing long-term consequences, and running up against limits may provide the

frustration that gives the experience of otherness its virtue-promoting qualities (172).

Furthermore, the otherness of games is not a real otherness, but only a kind of humanness:

“computer games often provide experiences in which we are everywhere surrounded by

ourselves – our own creations that are controllable, instantly disposable, and re-creatable”

(172). Digital characters then, are not ‘other’ enough to be worthwhile for Briggle. They are

not properly non-human because they have been designed by humans for humans. For Briggle

the nonplayer character can never be recognised as a being, both because the player is always

too powerful to be able to recognise it, and because it has none of the true alterity of the

natural world. The power dynamics seen in games, then, are inevitable, because reconciliation

with a true other is not possible.

Plumwood argues that in order to relate well to nonhuman others, things other than

rationality, consciousness or ego must be recognised as comprising their personhood. Game

characters are not reasonable or possessing of adequate ideas, as Spinoza would put it, but

they do have logic and intention. It’s often a murderous intention, but it’s a clear one. And

intention, Plumwood suggests, could be a contender for a recognition of personhood not

based on similarity to humans. As Bates has noted, a fictional character having a clear need

also helps humans relate to them better. Plumwood acknowledges that “full intersubjectivity,

requiring the mutual recognition of sameness and difference in dynamic interplay and the

sharing of similar conscious states, may only be possible between humans and humans for

some states” (213) but observes that there are other sorts of mutual interaction that occur

between humans and creatures with no linguistic (babies), subjective (plants), or dignified

(kittens) ways to relate. The way humans can most appreciate and relate to others requires

holding the self and other in balanced tension; “the resolution of dualism requires, not just

31

recognition of difference, but recognition of a complex, interacting pattern of both continuity

and difference” (67).

The affective structures that exist between humans and nonhumans in reality are paralleled in

relationships between humans and nonhumans in computer games. The same logic of

dominance that characterises capitalist interactions with the natural world is replicated in the

attitudes players are encouraged to take towards game environments. Players’ treatment of

nonplayer characters mirrors how master/slave dynamics work in real contexts. It’s nice that

Western civilisation has finally found something to subjugate that isn’t morally problematic at

all.

HUMANS AND ROBOTS

Whitby suggests that there are ethical implications to consider when robots (including

incorporeal robots like NPCs) interact with humans in a human-like way. When robots

resemble humans in physical appearance, in behaviour, or in social role, Whitby argues,

humans have different expectations from them, and obligations to them (3). Even if the robot

is not sentient, even if it can’t experience suffering, even if it is the property of a human, and

even if the human owner knows it cannot suffer, as long as a robot is human-like in one of

these three ways it occupies a particular ethical place, and is owed some consideration. This is

especially the case, he asserts, when it comes to damage and destruction. The destruction of

ordinary objects, at most, represents a waste of resources (2), but human-like robots are owed

more grace due to their greater capacity to affect humans. Because human-like robots have a

greater capacity to affect humans more than other objects, Whitby argues, more care must be

taken in interactions with them so the feelings of humans aren’t hurt. But it also works for

pleasurable interactions too. One must be careful in one’s dealings with robots so that they

don’t affect you too enjoyably either (otherwise you might find yourself preferring them to

humans). In Whitby’s view, you must be careful with human-like robots, because they may

make you feel too much in either direction.

32

Coeckelbergh notes that humans generally do not think of robots as having moral status and

would have a hard time arguing that they have any kind of moral agency or that we have any

particular duties towards them. “Robot rights” seem silly to ascribe to beings that are “mere

machines” (The moral standing of machines 62). At the same time, however, people who

interact with robots often do just this. They act kindly to the robot, take into account its

perceived interests, and generally act politely. They ascribe emotions and intentions to the

robot, take care not to ‘hurt’ the machine, or even love and care about it (62). There exists a

big difference between what people think about the abstract moral standing of robots, and

how they actually behave towards robots in real life. Coeckelbergh argues that this is because

humans interact with robots not based on properties like sentience or sapience, but

relationally: a robot appears different to different people at different times, and its moral

status depends on its place in an existing web of social relations (66). He also notes that “We

already know the entity as and since we are already standing in relation [to robots]. There is

already a relation before we talk and think about the question of moral standing or moral

status. At the moment that we think about moral standing of the robot, a linguistic, social,

material, etc. relational structure is already in place, has already grown, which for example

make us think of the robot as a ‘machine’ or an ‘it’ rather than a ‘companion’ or a ‘he’.” (70)

[emphases in original]. This is because “Our personal construction of the robot is influenced

by the way our culture constructs machines, and this construction is not only a word process

but also a living process, it emerges from a living and changing whole we call ‘society’ and

‘culture’ … When we talk and write about moral standing, we do not start from a blank slate”

(69). Interactions with robots are coloured from the beginning by ideas about robots found

elsewhere. Opinions about whether robots are machines, toys, beings, worthwhile, interesting,

morally significant, etc., have already been formed in an individual’s mind, and can only be

confirmed or made complicated.

Kate Darling has demonstrated that robotic toys can be the object of genuine adult sympathy

and affection. In an experiment designed to discover the limits of human empathy, groups of

adult humans were given fluffy robot dinosaur toys to play with, and then after a period of

time were told to torture the robots to death. Darling found that the humans were unwilling to

harm the robots, even though they were aware the robots would not suffer. This effect was so

pronounced that even bribing the participants with money and promising that their own

33

robots would be spared if they destroyed another person’s robot were fruitless. It was only

when the participants were told that all robots would be put to death if a sacrificial robot

dinosaur victim was not chosen that the humans acted to destroy a robot (Darling 12).

Although group dynamics, disapproval of material waste, and the toy’s fluffy materiality are

factors that do not apply to single-player computer games, Darling’s research suggests that

humans can very easily sympathise with artificial beings that are obviously artificial and can

suspend disbelief (see also Duffy and Zawieska on this point) to relate to robots.

In a less playful context, the performance artist Verdonck creates machines that work

themselves to death. These machines are not representative of living beings – they look like

car parts, they move jerkily, they have no capacity to sense the world or make sense of it. In

Verdonck’s Dancer series, they dramatically whirr, flail and grind themselves to obsolescence,

their purpose in life the cause of their death, their work working against them. Their only

function is to destroy themselves. As they break down with metallic screeches, wobbles and

smoke, their plight seems somehow tragic. They are the inevitable cause of their own

downfall. These are not humanoid robots – one of them is literally a car engine on a plinth. But

their drawn-out, ironic demise, their human size, and their setting in an art gallery rather than

under the bonnet of a car make them objects of sympathy, encourage comparison and

reflection, and even appeal to narrative archetypes. Verdonck’s work shows that beings do not

even have to appear or function as living things in order to be affective.

The researchers above torture robots in order to elicit feelings in humans. They affect

machines in the aim of affecting humans. The machines’ capacity to be affected is itself an

instrument of affect. In this, they demonstrate that the meaningless suffering of machines can

be made meaningful when it is associated with ideas of good moral character (a truly nice

person couldn’t even harm a robot), made analogous to human situations (the robot fruitlessly

working itself to death seems familiar), or call on archetypes and narratives (the robots’

programming leading to its doom seems Oedipal or Tragic). So while a human-like appearance

and even function is not required in order for a being to appear affected and be affective, a

human-like story or role is. As Spinoza tells us, “by the very fact we conceive a thing, which is

like ourselves, and which we have not regarded with any emotion, to be affected with any

emotion, we are ourselves affected with a like emotion” (147), and as the researchers and

34

artists above tell us, what we consider to be ‘like ourselves’ does not have to be a human

appearance or behaviour, but rather a familiar story.

Humans do not have to recognise things as people in order to have personable interactions

with them, these researchers argue. Exploring the limits of this ability, Nash and Verdonck

show that people have affective interactions with things they know perfectly well to be

mindless machines. Nash shows that the affective machines don’t have to have a physical body.

Verdonck shows that affective machines don’t have to be human-like. The experiments these

researchers run ask participants to consider machines as things that can be affected, and show

that the nature of that affection does not have to be novel or complicated in order to be

moving to human observers. Nor do the mechanisms have to be groundbreaking—Verdonck’s

machines perform ordinary machine functions, and Nash’s machines only require the clicking

of a mouse. Clearly humans can feel things toward and about robots, even if they are presented

as ordinary, unfeeling and virtual. Setting obviously helps. These affective interactions took

place in an academic conference, an art gallery, and a virtual sculpture garden in Second Life.

Materiality and the presence of other humans certainly has an impact. But these

considerations do not illegitimate the reality of robot-human affective loops, only help to point

to the circumstances in which they may be best instantiated or intensified. The conclusion

from these researchers is that robot-human relations may be best facilitated in a white space

environment in which classic narratives are invoked and some peer pressure and emotional

blackmail occurs.

However, as Coeckelbergh points out, perception and expectation play a large role in how a

human approaches a robot. Nonplayer characters are different to other kinds of beings, and

subsequently relationships with nonplayer characters are different to relationships with other

kinds of beings. And because of this the relationships are often classified as less important,

less serious, less deep, less durable, less real, and more potentially damaging. The perception

of human-NPC relationships as inferior to other kinds of relationships creates a cycle whereby

less is expected from human-NPC relationships by both developers and players. As

Johansson’s game developer interviewee says in terms of developing artificially intelligent

nonplayer characters, ‘“[there] is always a black hole of AI, where you are programming an

incredibly thoughtful, clever, brilliant AI behind the scenes and the player still reads it as a

35

scripted interaction”’ (211), implying that there is little point in programming complex

behaviour because players will dismiss it out of hand. Because of the perception of nonplayer

characters as dumb and programmed for limited functions, if nonplayer characters became

sophisticated and clever (as researchers such as Arrabales et al; Creed and Beale; Dignum et

al; (Guckelsberger et al.; Holtgraves et al.; Huang and Terzopoulos; Kasap et al.; Korn et al.;

Magerko et al.; Martens et al.; McDonnell et al.; Morrison and Martens; and Scholl and

Tremoulet; are attempting) no one would really notice.

FREAKS AND NORMIES

Much of the research about human-robot affections eventually ends up contending with two

speculative futures: what if interactions with robots lead to people becoming more machine-

like themselves, or what if robots become so lovely that people start to prefer them to other

people? Or, as Coghlan argues, are the ways humans treat robots entirely inapplicable to how

they treat other things? (Sex robots are the apotheosis of these concerns, and the ethics,

legalities and weirdness of their use is much discussed in robot philosophy.) These

discussions of the unintended results of technology are not new; literature is full of

speculative tales of human-robot relationships from Pygmalion to The Machine Stops.

In terms of computer games, these two scary futures speciate into: what if people start

treating humans or animals like they treat like nonplayer characters, or what if people start to

prefer fantasy people over real people? To which the answer is that

player-character/nonplayer-character interactions are only one example of a mastery mindset

already evident in many areas of life, and reality already contains plenty of fantasy. Perhaps

the scariest future is the one that is happening: games are part of what keeps the world

chugging away as it is.

Thankfully, queer theory has a lot to say about non-normative attachments, relating differently

to others, and the problems of liking someone you shouldn’t be liking. Mostly that it’s pretty

great. The twin dangers of human-robot affections, that it might result in everyone being mean

or everyone falling in love with the wrong object, are dangers to the ideal of everyone happily

36

getting along and falling in love with handsome young princes. But these are ideals that never

described reality. How humans interact with computer game characters reveals not only the

social norms that are programmed into games, but also norms that exist between humans.

Affective relationships with computer game characters are not normative, since their

fictionality is supposed to dissipate any strong feelings (Tavinor). But humans have feelings

towards things that don’t exist all the time (Spinoza 254), so attachment to computer game

characters is nothing out of the ordinary.

There are two ways queer theory can be used in this thesis. The first is to point out that

relationships between humans and nonplayer character, although an everyday phenomenon,

are still often seen as somewhat corrosive to individual character or the fabric of society. The

second is to identify the norms by which human-NPC relationships operate, and discover what

is being achieved by those rules. This thesis will mostly take the second approach. There are a

number of ways to observe norms. Happily, a number of manuals have been published by

game designers and researchers which proscribe nonplayer character behaviour and the whys

and hows of their existence. Where NPCs are praised or criticised by designers, reviewers and

gamers, patterns can be distinguished in what constitutes a good or bad interaction.

Observations can be made of games that typify human-NPC relationships, and those that

attempt to do something different. Examples can be found of NPC behaviour that is wrong,

strange, glitchy, or otherwise contradicts expectations. Queer theory in this sense is concerned

with what is normal and what is abnormal, and how these things are coded into games.

There are things that players are supposed to do in games. Players are supposed to defeat

things, acquire things, escape things, move things, maintain things, be things, do things. These

things are clear, spelled out in tutorials, and reinforced by material rewards, punishment

(including death), NPC dialogue, and limits. Not knowing what to do in a game is widely seen

as a sign of bad game design. So most of what a player does involves a certain should-ness,

supposed-to-ness, intention, paths, and norms. And this is where queer theory can be helpful.

PHILOSOPHERS AND PHILOSOPHERS

37

Plumwood and Spinoza are a somewhat contradictory pair to guide a thesis. They differ

greatly on the relative worth of women and animals (see Sharp 2011; Sharp 2019), and while

Spinoza comes to an understanding of the individual as a kind of function of the cosmos,

Plumwood only concerns herself with broad cultural milieu. However, Sharp notes that

Spinoza’s explanation of the construction of the individual resonates with contemporary

feminist structural analysis (2011). Sharp observes that “One of the implications [of Spinoza’s

idea of the individual as part of an overall mass] is that, if we want to fight domination, we

cannot change ourselves independent of changing the character of the mass to which we

belong. We need to act on what acts on us, in order to feel and live differently.” (Sharp 2011).

This is where Spinoza and Plumwood meet. Plumwood’s big picture analysis can also be

connected to individual affect. Sharp notes that “consciousness raising aims to reveal that

what feel like personal defects can be explained by larger patterns and structures. What

appears as an anomaly – “what’s wrong with me?” – is better understood as the product of a

pattern, a rigged system” (2011). Individual affect can be examined to show how it is related

to larger patterns, and larger patterns can be examined for their effects on the people subject

to them. For both philosophers, individuals exist and are constructed by a web of relations.

Haraway, too, argues that “beings do not preexist their relatings” (6), and Briggle also defines

the self as a relational self (155). Plumwood’s approach assists with identifying structures in

games that set up certain affective outcomes, while Spinoza can help with describing what

happens once the affect is occurring.

REALNESS

Realness is something that can trip games researchers up. One idea is that since games are not

real, they should affect humans in special fictional ways, affect them in lesser intensity, or not

affect them at all (Aarseth 50; Sageng 221; Tavinor). The debates about whether games are

real are often really debates of whether games are good quality, enjoyable, dangerous, or

worth taking seriously. But realness in games often operates in ways more akin to a drag race

than an ontology. Affect theory helps sidestep the issue by seeing all things as equal

participants in affect, just with different specialities. Deleuze’s interpretation of Spinoza

illuminates this when he states “animals are defined less by the abstract notions of genus and

38

species than by a capacity for being affected, by the affections of which they are ‘capable,’ by

the excitations to which they react within the limits of their capability … [Spinoza’s] Ethics is

an ethology which, with regards to men and animals, in each case only considers their capacity

for being affected.” (27, emphasis in original). It’s not so much of a stretch to say that this

statement about men and animals can be applied to women and game characters too. Games

are best thought of as things that can be affected, and affect, to different intensities, and this

cannot be generalised into types but remains specific to each encounter. For example, to my

laptop Portal 2 is an intensive program that makes it overheat and shut down, to me it brings

back memories of a long-distance relationship, and to my cat it is meaningless sound. The

affective capabilities of a game character depend on the objects involved, and cannot be

generalised. This means that while realness has a part to play in how intensely some humans

are affected, it is not the be all and end all of affect. Humans are affected by game characters

whether they think they’re real or not.

Game designers often argue that a player’s emotional journey should be at the crux of game

design – that game designers should centre the management of a player’s emotional state in

all their decisions (Schell 19; Chen; Isbister 1). In this view, a computer game is a kind of

ephemeral pinball machine that flings, rolls and bounces the player’s feelings around, all

assets working together as complementary cogs in an affective machine. The vast majority of

games researchers cited in this thesis argue that nonplayer characters should be more

affective, and they can become more affective by being more believable. This assumption is not

unfounded, at least when it comes to believability and player enjoyment (Bates), but the

concepts of believability, flow and immersion are often treated as synonyms in game

literature, and player enjoyment (or sometimes engagement) as the only affect that matters.

The idea that more believable characters are more affective asserts that affectiveness and

believability have a linear relationship, and rests on the assumption that mimicry of

something known to affective can produce similar affect as the thing being mimicked. The

more nonplayer characters accurately mimic something that is already known to be affective,

the more affective they will themselves be. Since humans greatly affect other humans, the

more nonplayer characters appear to be human-like, the more players will presumably feel

towards them. ‘Believability’ has a number of different definitions in different disciplines (Lee

39

and Heeter), but they all describe methods of making something that is not human appear

more like a human.

For many players and theorists, affect in digital spaces is modified by the fact that such spaces

are not other, more valued spaces, and actions are not other, more powerful actions. That is,

knowing that you’re playing a computer game is affective in itself. The responsibility for this

dissatisfaction is often placed on the shoulders of nonplayer characters. Briggle writes that “an

avatar in Mortal Kombat is not really the sort of thing that can suffer and die. Rather, the

important ethical consideration is whether and how participation in or exposure to such

alternate realities influences primary reality [i.e. real people]” (159). In this, Briggle is arguing

that the simulated suffering and death of digital characters is not a moral concern – the real

question is whether watching it has an effect on human players and observers and their

subsequent behaviour. This question of whether games change society or whether games are

products of society are not too difficult when viewed through an affective lens, which sees

classifications of internal and external, and boundaries between objects, as a bit nonsensical.

Briggle sees games, their content and forms as “products of wider social dynamics, which in

turn are influenced by games themselves” (160). Here he makes a case that games are not

outside the boundaries of real life, but instead (re)produce social dynamics, and are

(re)produced by them in turn. He notes that games are emotionally affecting, can seal or end

friendships, can challenge the player with morally charged scenarios, or build social

reputations. They can also be an arena in which important character traits like

sportsmanship, grace and persistence, are formed. For him, it is inevitable that playing games

will have some effect on the character of human players. As games as a cultural artefact

“embody a range of values and beliefs” (163), they “contribute to the character of culture, the

“soil” in which we find ourselves, thereby influencing what we do and how we think, or in

short who we are [emphasis in original]” (164). As part of culture, games influence culture,

and so their affective capabilities are not limited to individuals but the world more broadly,

though this thesis will mostly concern itself with smaller and more immediate affective

phenomena.

During play, the player’s body exists in relation to the digital character. The player has hope for

the final battle, frustration from the ally’s unhelpful actions, anger at the enemy, sympathy for

40

the slain deer, confusion in the shop, joy during the chase. For example, when a player of

Horizon Zero Dawn controls a player character in Horizon Zero Dawn to chase down a wily

robot dinosaur, that player will likely tense up, attend intently to the robot, ignore peripheral

details, become acutely aware of the player character’s response to input, quickly evaluate

opportunities and weak points, strike when they feel the chances of success are high, and feel

release and satisfaction. Their release is social (the player does not have to attend to this quest

any more), and physiological (they can relax). Through this sequence, the player is thinking,

feeling, acting and attending in relation to the game, and the game is acting in relation to the

player in turn. Game and player engage in an affective cycle in which each participant’s

internal state is affected by affecting, and is affecting in its affection. As Spinoza notes, the act

of receiving a desired affection can change the internal state so that the body wants something

completely different (155).

In their influential book The Media Equation, Reeves and Nass argue that “human responses to

media are social even though media supposedly offer only symbolic representations of people

in the real world” (46). They argue that humans unconsciously react to objects like computers

and places like digital virtual environments as if they were real people and places (81). This

results in people doing interesting things like talking to their computer, being polite to NPCs,

or feeling friendlier to NPCs that have some commonality with them. Reeves and Nass argue

that this reaction is automatic, unconscious, and unavoidable: “Everyone expects media to

obey a wide range of social and natural rules. All these rules come from the world of

interpersonal interaction” (5). Seemingly, humans will automatically socialise with anything

that seems even vaguely like it might hint at having a smidge of personality (82). Reeves and

Nass suggest that humans act contrary to their knowledge that they are dealing with fictional

scenarios: “media experiences are emotional experiences … this is true even though people

may seem to approach mediated presentations with the detachment necessary to render

pictures and words emotionless” (136, emphasis in original), and “many people will say that

media characters aren’t like other people, but people will often react to them and interact with

them in exactly the same way as they would with real people” (85, emphasis in original).

However, the observable tools and strategies used by people to interact with other beings

might not be a perfect guide to what they feel inside. Players may converse with a character in

41

a game not because they really think it’s a person but because talking is required to progress

in a task.

People behaving socially towards objects isn’t a problem in Spinoza’s view, but rather proof of

the efficacy of imagination. In his Ethics he states “If the human body is affected in a manner

which involves the nature of an external body, the human mind will regard the said external

body as actually existing, or as present to itself … the mind is able to regard as present

external bodies, by which the human body has once been affected, even though they be no

longer in existence or present” (98). When a human is affected by something, they

immediately imagine an external body that is affecting them, and so emotion (as a

consequence of affect) is always about an object. As Nash notes, “if we take a Spinozan view of

affect, then we do not need to be concerned with the difference between nature and artifice,

and can concentrate only on bodies’ capacities for affecting and being affected, defined by

their ‘compositions of relations.’ For Spinoza, a body can be anything that is capable of

affecting or being affected” (Nash 2012 18). Things get exciting though when a human is

affected by more than one thing at a time, since “if the human body has once been affected by

two or more bodies at the same time, when the mind afterwards imagines any of them, it will

straightaway remember the others also” (100). Associations form in the mind between objects

in the mind, such that “association arises in the mind according to the order and association of

the modifications of the body … every man will follow this or that train of thought, according

as he has been in the habit of conjoining and associating the mental images of things in this or

that manner” (101). Representation then, is associations of ideas that have arisen from

previous experience of being affected, and although the ideas may have nothing logically in

common or be held by others, they are bundled together in memory. So we come to the

conclusion that dill is nauseating because it was once consumed during an illness, or the

companion cube is lovable because it coincided with your victory, or pixillated barrel is

worrying because it lost you the game last time.

According to Nash, affect cycles can operate in digital realms just as effectively as in non-

digital ones. Humans are often at the centre of digital work: it is made for them and made

convenient for them. But although the digital centres humans, digital entities can and do affect

each other unseen by humans. In considering these interactions, Nash notes that “it is useful

42

to be able to examine the affective power of digital entities separately from the emotional

response it may elicit in human users of the work” (Nash 2012 19), as discussions of how

digital entities interact usually turn to how humans feel about it in the end. When data is

modulated into a display for a human that it comes to exist for that human, but data can also

be perceived by other digital entities. And when data is perceived, ‘modulation is constantly

occurring’ (5), that is, something is changing from one thing into another. Even when there are

no humans around, digital affect is happening.

Leino argues that emotions are better understood as structured relationships instead of mere

reactions, and that affect cannot be thought of as a series of isolated incidents but rather

permeates the whole game experience continuously. In his critique of game design discourse

on the need for more emotional moments and characters in games, he argues that “the sought-

after ‘emotionality’ of games, is not an unambiguous property of an object in the world like

colour or shape, which can easily be designed onto products. Neither is it … something

existing only in the mind of the player. It seems fair to suggest that the emotionality is a highly

subject-dependent property of a game; different players will have different emotional

experiences with the same game” (113). This echoes Spinoza’s observation that “Different

men may be differently affected by the same object, and the same man may be differently

affected at different times by the same object” (163). This means, Leino suggests, that “for the

study of emotions in play … the extremely emotional moments cannot constitute the whole

area of interest, which, instead, should be the emotional side of the player’s mindset in

general” (114). Although game designers may love to think and talk about passionate, fully

‘immersed’ players crying in dismay or roaring with rage, it is the player’s overall moods,

orientations, and attitude in the whole of gameplay, even in the quieter, more banal moments,

that must be considered in a study of game affect. Noting that “more emotional games are

often dubbed as the next milestone in the development of computer games” (113), Leino

argues that most game designers are going about emotional gameplay in the wrong way. He

asserts that extreme emotional events cannot be designed to hit a player with the right

intensity at the right time – instead, these events are the result of cumulative interactions and

the development of solid relationships.

Discussions of realness takes as their premise that nonplayer characters are unbelievable or

unreal to begin with, which, as Plumwood would argue, is already assessing them in terms of

43

their lacking compared to humans (115)). It is tied to the concept of immersion, the idea that a

game experience is more intense (and therefore more meaningful, enjoyable, memorable, and

everything else), if a player gets swept up in events and forgets, on some level, that what they

are experiencing is fictional. If the player remembers that the nonplayer character is a being

animated by code and not by a soul, they are less likely to take it seriously, find interactions

meaningful, empathise with it, or enjoy its company, Warpefelt and Verhagen imply. For much

research into the affective capacity of nonplayer characters, whether a digital character is

realistic or believable is assumed to be very important to the feelings people develop around

it.

The centring of believability and investigations into its lack also generates the opposite

problem – the worry that some people may find games and characters too believable,

especially compromised groups such as children, the mentally ill, the intellectually disabled,

the lonely, or the gullible. Suddenly, the stakes become high. A character that is too believable

may become a dangerous Pygmalion, luring vulnerable people away into delusion. This

paradox of fiction is what affect theory neatly sidesteps, by acknowledging that whether the

quintessence of a thing is ontologically solid or not, it can still engage in a relationship.

YOU SAY ARCHETYPE, I SAY CLICHÉ, LET’S CALL THE WHOLE THING OFF

NPC’s “provide many functions to players” (Warpefelt and Verhagen), “provide an important

service in video games” (Warpefelt), “fulfil companion duties” (Guckelsberger et al), are

“designed in support of [the] goal [of] … helping the player learn something or have an

emotional experience” (Martens et al.), and communication with them is “initiated and

controlled by the player” (Morrison and Martens). Overwhelmingly, NPCs are seen by game

designers as mechanical providers of a service, much like a vacuum cleaner or a vending

machine. Like a hungry commuter may press several buttons in order to receive a particular

chocolate bar, a game player may click through a dialogue tree in order to receive a particular

item, piece of information, or quest.

To achieve certain desired player affects, game designers often choose to represent beings that

elicit these behaviours and feelings in other contexts. In this, they rely on reputation, cultural

44

knowledge, common sense, and other things that are ‘intrinsic’, ‘instinctual’, ‘automatic’ and

‘obvious’. Originality of characters, in this sense, is a burden, because if a player doesn’t know

what a character is, they don’t know how to act. Immediacy and predictability is prized by

designers, as speed is compelling and enjoyable. Discovering the proclivities of each character

and negotiating relationships requires players to stop and think rather than immediately

react.

A few game designers have argued that digital characters are necessarily disappointing due to

the very nature of the computer game. Gonzalo Frasca notes that “video-game characters are

flat for functional reasons: what is important is to get the plot moving forward. The question

that needs to be answered is ‘what happens next?’ and not ‘why the character behaved in such

a way’” (168). Because games are about action, the depth of non-player characters must

sacrificed in favour of drama, Frasca argues. In noting games’ differences to literature, Raph

Koster also makes the point that “Games are good at objectification … Games tend to quantize,

reduce, and classify … Games are external – they are about people’s actions.” This is in

contrast, he says, to stories, which Koster argues are about empathy, feelings and subtlety.

According to Koster, games are about the player’s feeling of mastery, and this necessarily

results in shallow relationships between players and non-player characters. But as we shall

see, this is not always the case.

WHAT’S YOUR TYPE?

Isbister says that there are 14 relationships a game character can have with a player: as

minions, rescuees, pets, sidekicks, allies, guides, mentors, obstacles, enemies, competitors,

boss monsters, arch-enemies, audiences, and hosts.

Warpefelt says that there are 12 services that game characters can provide to players: sell and

make stuff, and provide services, provide combat challenges and provide loot, provide

mechanical challenges, give or advance quests, provide narrative exposition, assist the player,

act as an ally in combat, accompany the player, make the place look busy, and be vanity items

(pets).

45

Dominguez et al. have shown that players act according to roles even when there are no

mechanical restrictions upon their potential actions. This indicates that narrative can act as a

constraining device, even when the narrative is as thin as a single sentence (“you are a thief”),

or is generated by the player themselves (‘headcanon’). Their study indicated “that players are

strongly consistent with their roles, regardless of whether their role was explicit (assigned or

chosen), or not” (3446). Player character and player roles affect affordances (von UexkuH ll’s

‘tones’), that is, what the player can perceive what objects (including themselves) can possibly

do, and roles also influence the range of actions that are appropriate for a character to

perform. For example, it is more narratively appropriate for a thief to sneak into the back of a

building than fight their way through the guards at the front, as the idea of thief carries with it

connotations of sneakiness, judgement and speed rather than recklessness, savagery, or

mayhem. Even in the absence of concrete rules to enforce this, players will limit themselves to

preserve coherence in their roles. If a player is told or chooses to begin to play a game as a

thief, they will continue playing the role of the thief even when there are no rewards or

punishments associated (3447). In short, player roles influence how a player will act in a

game, and players pick up roles extremely easily. While the expectation of gaming sets up an

orientation towards excitement and action, player roles narrow the orientation further,

towards particular actions. As Dominguez et al. note, affordances are not only physical limits

on the actions the player can take, but mental ones on what the player imagines can be done.

Dominguez et al.’s research suggests that players very often adopt a persona that limits their

own choices (3439), even when the game allows a broader range of possible actions. A

player’s adoption of a role results in “a significant relationship between their role and their in-

game actions; participants role-play even if not instructed to, exhibiting a preference for

actions consistent with their role.” (3438).

Roles are often considered an essential way to impart information to the player quickly and

implicitly about the kind of behaviour they can expect from a character and the actions they

should take in response (Isbister; Reeves and Nass; Freeman; Warpefelt). Roles are an

abbreviation, a short-cut, an unavoidable, unconscious dip into a deeper pool of knowledge

that the player already carries inside them. They mean that designers can draw from a well of

pre-formed associations, assumptions and expectations instead of exerting effort to create

46

unique characters and teaching players how to interact with each one individually. Reeves and

Nass note that “‘Simple’ also means predictable. … When people know what to expect, they

can process media with a greater sense of accomplishment and enjoyment. The desire for

simplicity can be pernicious however – people stereotype media by gender with little

encouragement” (254). Yet, for all their simplicity and straightforwardness, game characters

do not often elicit an intensity of emotion towards them as beings. Instead, they facilitate

affect by standing in for concepts with which the player has an existing relationship.

Reeves and Nass argue that squashing people into boxes is not only inevitable, but highly

pleasant for the squasher: “categorisation simplifies interactions … you can make assumptions

about how he or she will talk, think and behave, and how you should respond. The

predictability gained by applying a label makes people reluctant to scrutinise the label’s

accuracy” (144). Categorising others reduces awkwardness, embarrassment,

misunderstanding and conflict by giving people verbal scripts to follow, forestalling

uncertainty about other people’s unknowable inner worlds, and eventuates in a sense of

achievement when people play their roles successfully. This suggests that game characters are

actually the epitome of the human urge to judge and assign qualities to others – game

characters are made to be easy to categorise, easy to predict, and interactions with them are

straightforward. They are especially easy to categorise as ‘good’ or ‘bad’, a basic and

immediate occurrence, Reeves and Nass argue; “Brains are hard-wired to distinguish between

good and bad as efficiently as possible” (112). In making enemy NPCs spiky and friendly NPCs

cute, game designers are merely taking to the extremes psychological constructs that

seemingly already exist in the human mind.

Isbiter (a student of Nass) makes the case that good character design comes from

understanding the basic psychological principles that underlie interpersonal interactions,

rather than assuming that a special kind of relationship forms when humans interact with

digital characters. For her, players’ use of everyday social conventions in novel digital

situations means that study of everyday social conventions can be applied to interactions

people have with digital beings. Furthermore, Isbister contends that knowledge of these

everyday conventions can help game designers manipulate people into believing that NPCs are

something worth bothering with, without them even being consciously aware of it. This is

47

achieved mainly through the appearance and the animations of the characters, she argues.

Schell also argues that appearance of game objects should be closely tied to their behaviour:

“if two objects behave the same way, they should look the same. If they behave differently, they

should look different” (136). This link between appearance and behaviour is a way to manage

relationships between the player and the object. Appearances are tied to what the player can

expect from the object, and if the object is familiar, the player can immediately recognise it and

does not have to renegotiate a new relationship.

Roles do important work for the player, defining the relationship the player will have with the

character before they even meet them. The designer takes care of all the evaluating, testing,

learning and judgement that would usually have to take place when a person meets another

being. As Isbister states, “social roles are, in a sense, libraries already present in the mind of

the player that a designer can tap into to create satisfying interactions with NPCs” (227). And

it means that in order to make the player feel something, the designers just have to make a

character that expresses the corresponding traits, e.g. to make the player feel submissive, they

create a domineering character. Isbister makes the case that roles “enhance the chances of

stronger emotional experiences for players and create a more game-play-integrated

experience of NPCs” (225) by “help[ing] people to engage in interaction with others without

having to negotiate everything about how each person will act, and often, without even having

to get to know one another very well” (226). She argues that this is essential in the time-poor

context of a computer game, where the player would rather be off doing something

interesting. She admits that “social roles can be barriers to greater intimacy, but they are very

helpful in everyday situations in which a person has short and targeted engagement with

another” (227). It is the shortness and task-based nature of interaction with NPCs that results

in designers’ need to work with cultural short-hands like roles and stereotypes. If speed and

goal-achievement is the criteria for interaction with games, roles are a requirement for

smooth interaction, as confusion, embarrassment and conflict will invariably occur when the

player doesn’t know exactly how to act (Isbister 226; Lankoski, 314).

These psychological perspectives of game characters elucidate the processes that go into

judging a digital character, and the conventions that form the basis of interactions with them,

but rely on a fixity of identity and relations that suggests that human interactions with digital

48

characters can only ever be shallow and exploitative. They argue that typecasting is inevitable,

and suggest that a way to make better relations between humans and digital characters is to

further entrench those types, to play further on humans’ immediate judgements of cute or

threatening. Any disappointment arising from interactions with digital characters is because

they fail to fully inhabit their intended archetype. But the danger of this view is that it

positions simple, knowable, discrete roles for digital characters as unavoidable, and existing

roles of ally, enemy and so on, as inevitable results of some universal human need for

categorisation. It suggests that a certain affective state can be definitely achieved if designers

follow a certain set of rules. But this ignores the historical contingency of current character

roles.

Some, like Mackenzie Wark, have argued that the shallowness of digital characters is a good

thing. Familiarity with artificial worlds, Wark suggests, empowers players to discover the

artificiality of capitalist reality. Instead of sucking players in to a shadow world, games may

reveal to the cave-dwellers the true nature of the shadows dancing on the walls. Shallow

relationships with digital characters reveal the shallow ways capitalist subjects are meant to

interact with each other, and so are secret liberators through their frustrating fakeness.

Briggle notes “game-worlds can be fruitful places to try out alternative identities, thus

potentially increasing one’s empathic capabilities” (169), so suggesting that games, rather

than being sites of hegemonic enculturation, may be spaces in which individuals can perceive

and critically evaluate underlying rules and so increase their overall capacity for seeing things

in new ways.

INTERTEXTUALITY

Lankoski and HelioH illustrate how character design helps to define gameplay as a whole, and

how genre conventions, including character design, shape player expectations and subsequent

play styles. “From the player’s point of view,” they argue, “action is an important feature of a

game. Action can be created and directed by raising expectations, building up motivations and

setting goals for the players. An important tool for setting up motivations and goals for the

player are well-defined characters with distinct natures and needs” (311). Characters create

49

conflict, and conflict requires action (312), so good characters are important for exciting

action. The goals of characters become the goals of the game.

They observe that games are highly intertextual (or paratextual (Beavis et al. 33)) in that they

rely on previous games and dramatic conventions to quickly communicate information to

players about the nature of characters and settings. This has implications for how players

perceive the affordances of a particular game: “games are not only read in relations to other

texts but the reading also affects the choices made by the players. Players usually behave

according to the rules of the genre and they use patterns of behavior learned from other

sources while playing (i.e. sources to which the game refers)” (316). And it also allows games

to get away with characters who are quite flat in themselves but are fleshed out by references

to other characters in other places (316). A large part of a player’s knowledge of a game comes

from sources other than the game in front of them. As Bateman notes, this also has

implications for how designers design games: “the habits we pick up when we are playing

games that go on to assert constraints upon us when we make games. To understand games in

terms of player practices is to appreciate that anyone who thinks their game concept was

entirely their own invention is fooling themselves; we don't invent new games from whole

cloth, we iterate on the player practices we already learned and crossbreed them with new

influences, often from other kinds of media.” (Bateman ‘Game Dissonance’). Games use genre

and reference to re-present existing ideas, which has advantages both in terms of the player’s

speed of comprehension and the amount of work required to make a game. Games therefore

often expect players to have some familiarity with other games, and a game only often makes

sense in a context of similar media. Even the names of popular genres, like roguelite or

metroidvania, only make sense if one is familiar with an oeuvre. “The mental images of things

are more easily associated with the images referred to things which who clearly and distinctly

understand, than with others” (254), as Spinoza says.

Lankoski and HelioH note that players come to games with assumptions and expectations, and

this effects how they perceive the affordances of the game, and how they act in the game:

“players’ basic assumption that a game should be interesting and challenging affect the way

the game is played. Players are often seeking action even if it is against the nature of the

character they are playing. We use the term expectation of gaming to describe this

50

phenomenon” (315, emphasis in original). Players are often seeking to do things in games, and

seeking objects to do things to. This expectation is “also created by intermediality and

especially by genre” since “players usually behave according to the rules of genre and they use

patterns of behaviour learned from other sources while playing” (316). They suggest that

during the playing of a game, players are always referring to intermedial knowledge, that is,

we “understand any action in a narrative by our experience of similar action in other

narratives, and that our experience is an aggregate of details arranged in generic categories of

action – murder, rendezvous, theft, perilous mission, falling in love etc” (Fiske, qtd. in Lankoski

and HelioH 316). The game is made comprehensible and intelligible by familiarity with

convention.

This sense of games as an ongoing self-referential conversation rather than a collection of

individual media productions also makes it easy to talk about games and characters in

aggregate. This is both an asset and a trap. Though exceptions exist for any generalisation, and

more experimental indie games are being made now than ever before, the wide adoption of

conventions means that many commonalities exist between in terms of characters. Some of

these commonalities exist because they are easier to program. Some of them exist because

they are easier to teach to programming students. Some exist because of received wisdom

from game designers past of the right way to do things. Some exist because game designers

love games and want to make more things like the things they love. That said, no two games

are exactly the same in how they treat their characters.

The iterative nature of game design, both in terms of the production of a single game and of

this intermediality, means that the connection between player action and character design is

solid. Therefore a way to study computer game characters is to look at what actions a player

can possibly take.

AFFECTIVE CHARACTERS THROUGH MECHANICS

51

In comparison to visual design and animation, the affective capabilities of game mechanics

have not been addressed as thoroughly. Mechanics also have certain connotations for gamers

as the real, immutable, deepest part of a game that only real gamers appreciate, while art and

design are girlier, ephemeral, and don’t affect the core experience of playing (see Koster). It

makes sense then, that affect has primarily been studied as an effect of game art, as it is visual

cues that supposedly give games their mood. But affect can work in games through action and

movement of a body as well as visual impressions. The game designer Bithell also notes that

characters in games, like characters in action movies, are defined and differentiated entirely

by their actions, their “gameplay verbs” (Yarwood).

While they are widely considered important, what game mechanics actually are remains

under some doubt. Many designers and theorists offer different definitions. The game

designer Schell notes that mechanics are “mysterious” (129) but “the core of what a game

truly is. They are the interactions and relationships that remain when all of the aesthetics,

technology and story are stripped away” (130), while Koster sees them as “an intrinsically

interesting rule set into which content can be poured” (120) covered by “’dressing’ [which] is

largely irrelevant to what the game is about at its core” (84), Nyugen and Ruberg define

mechanics as “units of interactivity” (698), Winn argues that they are “the formal rules that

define the operation of the game world, what the player can do, the challenges the player will

face, and the player’s goals” (1024) Jagoda and McDonald define them as “the set of

techniques for interacting with a game world that are arbitrarily mapped through an interface

to player gestures and are both constrained and enabled by a game’s platform” (176), De

Gloria et al. see them as a kind of negative space: “the limits of the designer intervention on

how games actually take place” (295), and Sicart identifies multiple levels of mechanics, the

fundamental core of which are actions “(repeatedly) used by agents to achieve a systemically

rewarded end-game state”. Mechanics, then, are the bones of a game, the thing that remains

constant while the settings, characters, narrative, can change. And if the game’s mechanics

change or cease to exist, then the game itself changes or ceases to exist. “If there’s no core

mechanic, there’s no game at all”, Koster states (122) (see Bateman for a refutation of the idea

of fiction as wrapping paper). Mechanics seem central to the identity of a game.

52

While all these definitions are vague, they have some concepts in common. Mechanics are

about doing and motion. They are the actions the player can take, and the limits of those

actions. Mechanics define the boundaries of action, of what is possible and what is not. They

are, in short, the about the capacity to affect and to be affected.

As Lankoski and HelioH have shown, the design of characters influence player action, both by

having goals and conflicts of their own, and by referring to bodies of other work as guides. The

affective capacities of characters lies not only in their power to make players feel happy or sad,

but the power to make players replay a level, fly, drown, lose or gain items, or survive another

night; in short, to increase or decrease their own power of activity.

Mechanics are about how the player can affect the game world and its characters. But the

player’s position in relation to others and their role in the game shapes how the player will

affect nonplayer characters. Mechanics, then, also encompass the player as a relational being,

as an affective agent.

CONCLUSION

Because humans can be affected by ideas and rocks and all sorts of ephemeral and nonhuman

things, they can be affected by game characters. Even when a player knows that game

characters don’t feel emotions or have to live with lasting consequences, they can still be

affected by them. And even though game characters are made out of code and pixels, they can

still be affected by players and each other. Game characters are like any other sort of of entity

in that they can be part of an affective cycle.

However, the ways in which players are affected by characters and the ways in which

characters are affected by players are specific to games. Players and characters take particular

actions in games, and they take particular actions toward each other. Players feel differently

towards characters in games than they do towards characters in other media because they act

differently towards them than they do to characters in other media. The kinds of actions

53

characters and players can take in a game is necessarily limited, so a game tends to be defined

by these actions, and this has led to the development of various genres and conventions. This

has implications for the kinds of affective cycles that they can engage in.

The ways in which players act around characters in games has parallels with how people act

around each other. Players can feel obliged to be polite to digital characters and can feel

uncomfortable torturing robots in group settings. But games are also designed to make

players into very powerful figures in the game world, and the ways in which this is achieved

mirrors how power works in non-game spaces. Better characters could be designed if the

relative positions of game characters and players were investigated further. Realism isn’t the

only thing games can strive for, and striving for a simulation of what is already familiar short-

changes the uniqueness of game characters and ignores the possibility of new weird

relationships.

54

CHAPTER TWO: ANIMALS

INTRODUCTION

In the best article on computer games ever written, “The ethics of human-chicken

relationships in video games: the origins of the digital chicken”, Fothergill and Flick discuss the

history of the domesticated chicken, how uses and perceptions of it have changed over time,

and how representations of chickens in computer games typify a modern understanding of the

animal. While they make a good case that the representation of chickens in games as silly

overemotional foodstuffs is inevitable given their real-world commodification, the

straightforward connection between real chickens and digital chicken points towards the

necessary difficulties of distinguishing between othernesses.

Animals and NPCs are both nonhuman. This position in relation to humans creates some

commonalities through the ways in which otherness is constructed. Further similarities are

designed into animal NPCs. Animals NPCs have a lot of the same uses in games as animals do

in other contexts, and designers intend them to have similar relationships with players and

generate some of the same affect as real animals would. NPCs’ and animals’ common

otherness means that animal NPCs have a lot to say about how representation and affect work

together. Chicken NPCs aren’t a type of chicken, but chickenness is important to understanding

the digital chicken. The constructed similarities and differences between animals and animal

NPCs have important things to say about humans’ perception of both.

NPCs aren’t humans or animals, but everyone either wants or assumes players to have

relationships with them as if they were. Chickens aren’t some weird fantasy creature made up

for games; designers’ construction of them and players’ perception and use of them has some

basis in other media representations at least, if not real life. As Johnson and Verdicchio and

Coeckelbergh remind us though, animals and robots have some fundamental differences, and

one can’t easily be substituted for another. Robots and animals are different things, and games

55

are a different place. Not a completely separate place, a magical place, or an inconsequential

place, but, still, a particular place. Designers and media scholars might try to argue that

players behave towards digital chickens as if they were real chickens, but in the end players

are dealing with a very different beast.

This chapter is a little different to the others because it seems to define an NPC by their

appearance rather than their relation to the player. However, it will argue that ‘animal’ is a

relationship somewhat like ‘enemy’ or ‘friend’, animal characters afford specific affects that

human NPCs do not, and furthermore, animal NPCs are the epitome of games’ instrumentalist

relationship modelling. Animals can certainly be enemies or friends to the player, but they can

also be an other, and a study of their construction of otherness can help illuminate all

typologies of NPCs as political exercises rather than descriptive ones. That animal NPCs are

nonhumans representing another kind of nonhuman means that they have a lot to say about

the construction of the human in general, and the maintenance of the distinction between

actors and environments in games in particular. This chapter then, will include a short

discussion of the player-character’s self, whether it is human around animals, nonhuman

around other nonhumans, or nonhuman among humans, and tease out the implications for

relations with NPCs based on that contrast.

Animal NPCs have a basic relationship to players that is not affable or antagonistic: they are

resources. They exist to be used. Animal NPCs occupy a particularly interesting place because

players can have relationships with them that they cannot have with human NPCs. Even in the

nonhuman otherness of virtual spaces the ontological separations of human and nonhuman

are generally preserved. It would be a strange game that had human characters as transport,

food, pets or livestock, but a normal game that had horses to ride, rabbits to eat, dogs to pet

and chickens to farm. As animal characters serve different functions to human characters, and

players and animal NPCs can act toward each other in different ways to players and human

NPCs, the affective relationships they can have are different. The affect they can produce in

each other also depends on animals’ positions in other contexts. Animals, like NPCs, stand in

contrast to humans. They are seen to lack the human qualities of rationality, freedom, or

awareness of mortality or sense of self (Plumwood 43). In this sense, real-world animals and

game-world NPCs are in a similar position in relation to human players.

56

Animal characters in games bring up tricky questions about representation, and problematise

the idea that believability is a necessary quality for close relationships with game characters.

Because players have less experience with animals than other humans, and for many species

more experience with media representations than their actual bodies, it can be easy to

confuse the map with the territory. Players want different things out of animal NPCs than

human NPCs, and usually less. Like human NPCs, animal NPCs can either be multifaceted,

complex and affecting, or so functionally singular that it can be difficult to classify them as

characters at all. However, the design of animal characters takes these qualities to extremes

that human NPCs seldom inhabit. It is difficult to design a human character as lovable as

Fallout’s Dogmeat or The Last Guardian’s Trico, or as unmemorable as Skyrim’s rabbits,

Horizon Zero Dawn’s rabbits or The Long Dark’s rabbits. There is something available to

animal representations that are not available to human representations: an acceptance of the

fact they behave automatically, with no malice aforethought, or any aforethought, or any

thought at all. Their lack of sentience means that they cannot be reasoned with, or basically

responsible for their actions, only acted upon with either grace or violence. This is both an

asset and a curse to the creation of meaningful player-animal character interactions.

Apart from their role as a resource and particularly tricky example of representation, animal

NPCs also typify something about all NPCs. Everything an NPC is, an animal NPC is to excess.

They are a mish-mash of things and half-baked ideas about other things. They are background

decorations, or they are the most lovable thing about a game. Their place in the world is

especially determined by how a human wants to use them. A study of animal NPCs then, has

much to offer.

WHY PLAY WITH ANIMALS

ELVES

Defining the animal in games is not an easy matter, and is not just a case of anatomy. Humans

are certainly not the only humanoid creatures in game media, and games are filled with

57

sapient nonhumans. ‘Animal’ is a difficult category to sustain when so many games include

orcs, elves, goblins, aliens, spirits, gods, demons, dragons, robots, animate teddy bears,

ensouled swords, talking plants, and sentient rocks. The nonhuman is a familiar sight in

games. But while some characters look beastly, it is really their functionality that determines

whether the character is a ‘person’ to the player. It is the creature’s role in relation to the

player, not the presence of a snout or a tail, that defines a creature. It is not just a matter of

appearing human – personhood in games is a matter of relationships.

While many games represent animals that can be found in the real world, many others feature

cartoon animals, animal-headed humanoids, or mythical, fictional, extinct or rare creatures

that can never be encountered outside of media. Each of these depictions indicates a form of

intended relationship. Their body shapes are deterministic. While convenient for players, this

fixity is a warning sign that essentialism is on the way. Such a conception of identity obscures

how bodies are positioned as a result of forces and characterises power as a kind of natural

landscape that makes affect flow downhill rather than a mechanism that pumps affect through

particular routes.

There are important things to be said about nonhuman humanoids being positioned as a kind

of human racial category, but this isn’t the thesis for it. Poor gives a good outline of this topic,

while Higgins documents how “politically disruptive racial differences are ejected or

neutralized through fantastical proxies” (3), and Bjørkelo describes the way white nationalists

use this fixity of identity to confirm and celebrate their ideology. The rest of this chapter will

be concerned with how animal NPCs with some real-world equivalent are used in games to

generate certain affects.

NONHUMAN OTTERS

Animals and NPCs have one important thing in common: they are not human. This otherness

informs their affective capacity with humans. Otherness is both recognised and constructed; it

is capitalised upon to produce and maintain certain inequalities, Plumwood states. By not

58

being human, animals and NPCs are made common. A being is made into an other, Plumwood

argues, through a process of backgrounding, hyperseparation, incorporation, instrumentalism,

and stereotyping (53). These techniques will be discussed in further detail in terms of the

backgrounding of rabbits, hyperseparation of robot dinosaurs, incorporation of horses,

instrumentalism of dogs and stereotyping of chickens. The same things that are done to deny

animals’ entirety are present in the construction of NPCs, and come to a head in the

construction of animal NPCs.

Plumwood argues that nature is dehumanised (and humans are denaturalised) by it being

perceived and defined as mechanical; nature is “seen as non-agentic, as passive, non-creative

and inert, with action being imposed from without by an external force. It is non-mindful,

being mere stuff, mere matter, devoid of any characteristics of mind and thought. It lacks all

goals and purposes of its own (is non-teleological and non-conative). Any goals and direction

present are imposed from outside by human consciousness. The human realm is one of

freedom, whereas the realm of nature is fixed and deterministic, with no capacity for choice”

(Plumwood 110) Animal NPCs then, have no chance. Their digital bodies are doubly ‘an

instrument for the achievement of human satisfactions’ (Plumwood 111). Berger, too, points

out that capitalist relations with nature requires animals to be intellectually processed before

being consumed. The processed animal therefore creates the myth of the unprocessed wild

animal, who is more true and free, “an ideal internalised as a feeling surrounding a repressed

desire” (Berger 17). In terms of zoos, another place for encountering entertaining nonhumans,

he notes “The zoo cannot but disappoint. The public purpose of zoos is to offer visitors the

opportunity of looking at animals. Yet nowhere in a zoo can a stranger encounter the look of

an animal. At the most, the animal's gaze flickers and passes on. They look sideways. They

look blindly beyond. They scan mechanically” (Berger 28). While Berger perhaps overstates

his point, he has a point that animals in zoos are expected to act predictably, like robots, for

the enjoyment of humans. Given that Cartesian views of nature haven’t completely gone away,

resulting in ostensible similarities between animals and machines, designing animal NPCs

should be easy. However, animals in games, like animals in zoos, are disappointing because

they aren’t real others, and they don’t look back. Their alterity has been subsumed into a

comprehensible human system, and this can be cause for despair (see Epp-carter for a

thorough but narrow discussion on the negative affects of represented nature). But, as Chang

59

notes, this connection between nature and machine can be put to use too: “Games can offer a

compelling way to reconcile a deep connection to nature and the nonhuman world with an

equally important connection to technology and the virtual” (8). JanA ski concurs that “the

potential video games share with other works of fiction, i.e., to challenge the well established

preconceptions and explore new vantage points, should not be overlooked” (89). Animal

NPCs, as machine-like animal-like machines, are key to this opportunity.

Where Plumwood asks us to recognise what we have in common with animals without

incorporating them into humanness, Berger argues that the otherness of animals is necessary

to humans. He sees animals as a bridge between humans and nothingness and their similarity

and difference as essential in that role: “Animals interceded between man and their origin

because they were both like and unlike man.” (Berger 6, emphasis in original). The position of

animals, Berger concludes, also defines the position of humans. “The man too is looking across

a similar, but not identical, abyss of non-comprehension. And this is so wherever he looks. He

is always looking across ignorance and fear. And so, when he is being seen by the animal, he is

being seen as his surroundings are seen by him” (Berger 5, emphasis in original). Animals are

part of an environment, a sphere that encircles a person and is background to a human

foreground. When the background expresses some agency it reveals the environment not as a

set for human action but of a field of constant impersonal actions, of which humanness

emerges as just one situation among many. Dualistic conceptions of nature exaggerate and

essentialise difference and incorporate sameness, and this process can be seen in the way

animal NPCs are designed. This in turn, defines the player as a being in the game. Animals’

representation reveals what designers really think about the player and their place in the

world.

Talking about representation can get pretty confusing, pretty quickly. It is necessarily difficult

to distinguish between types of otherness, because, as Plumwood puts it, “Culture …

accumulates a store of … conceptual weapons, which can be mined, refined and redeployed for

new uses. So old oppressions stored as dualisms facilitate and break the path for new ones”

(43). Baker notes that the boundaries between representation and reality are fluid and

dynamic, as “human understanding of animals is shaped by representations rather than by

direct experience of them … representations do have consequences for living animals” (190).

60

Representation shapes how humans think about and perceive animals, and this affords only

certain affections (leading to disappointment at the zoo, as Berger points out). Plumwood

delicately pivots between talking about animals and talking about ideas of animals, while

Berger cheerfully leaps between animals, ideas of animals and representations of animals.

Both these approaches guide this chapter’s discussion of representation (Spinoza, as Sharp

(2011) points out, isn’t very good with animals). Animal NPCs can be compelling not only for

their apparent innocence but because animals and NPCs are both subject to the kinds of

power that shape human perception of nonhumans. In this way, humans already interact with

animals (at least unfamiliar animals) as if they were representations.

Animals have an existence which precedes animal NPCs. However, for many animal NPCs this

hardly matters, because what players know about them comes from other representations. As

DeMello says, “Animals’ physical identity is less important to their status and treatment than

their symbolic identification and their social meaning. In addition, how we classify animals

shapes how we see animals, and how we see them shapes how we classify them” (10). It’s

hyper-real maps all the way down, without any foundational territory. The actual differences

between animals and NPCs must be kept in mind. Johnson and Verdicchio argue that even

though robots are often compared to animals in terms of capacity and potential social role, the

two should not be conflated. They observe that although robots and animals share “otherness,

capacity to trigger humans to anthropomorphize and attach, trainability, and potential to

assist humans and harm humans. … when it comes to thinking about the moral status of

humanoid robots, legal liability, and the impact of treatment of humanoid robots on how

humans treat one another, analogies with animals are misleading. They neglect the

fundamental difference between animals and robots, that animals suffer and robots do not”

(292). Animals are affected differently than robots, and experience that affect differently too.

They exist independent of humans, whereas NPCs do not. Perceiving an animal as machine-

like is doing them a disservice, but perceiving an NPC as machine-like is just accurate.

However, the othering of animals and the bad design of animal NPCs is part of the same

phenomena.

WORKHORSES

61

Feinberg et al. write “Wherever we humans go, we seem to carry plants and animals with us”

(3), and this is true of digital spaces too. Animals in computer games are invisible in their

ubiquity. They provide introductory challenges in tutorial areas, populate planar forests, are

movable feasts of usable materials, and are the basis of whole genres. As well as being the

stars of the virtual pet, farming, exploring, zookeeping, and hunting genres, animals are the

unappreciated sidekicks of the RPG, the open-world sandbox, and adventure genres. They are

friends, pets, adversaries, transport, guides, collectables, workers and ornaments. What they

never can be is realistic. Instead of being realistic, animal characters are naturalistic, their

appearance and behaviour appealing to an idea of nature that seems truthy but are just based

on repeated media iterations. Through this, animal characters blow apart the distinction

between realism and believability that games researchers normally disregard.

Animals have at least twice as many legs to animate, fur and feathers to texture and render,

and incomprehensible or boring motivations. It seems strange that they should be

represented in games at all. Yet they have been there since the very beginning of games

(especially if aliens can be counted as a type of animal). Two early examples typify their roles.

In the 1975 text adventure Colossal Cave Adventure, the player encounters an enemy snake,

and uses a caged bird as a tool to frighten it away. In the 1981 platformer Donkey Kong, the

titular ape kidnaps the player-character’s love interest and spends the rest of the game

keeping her captive and chucking barrels at the player-character. These animals are enemies

and tools, but also conform to familiar narratives.

Animals are represented in computer games to replicate existing affective relationships, to

facilitate a simplified and intensified version of what is already familiar. As such, particular

animals are included in games for their existing association with particular affects, and

because of this animal characters as characters are often not very complicated. This is an

extension of the simplification seen in all NPCs, just turned to the concept of species instead of

social roles. In doing so, the concept of a species becomes a social role through its human use.

“How we classify rabbits—calling them pets, or meat, or lab animals—has to do with where

they live, and what they are used for” (45) writes DeMello, and what is true of rabbits is true of

NPCs. A pet bunny has different legal rights and a different affective relationship with humans

62

than a feral rabbit. NPCs are defined by what they provide to the player. The idea of rabbitness

translates to games through those roles. The aspects of animal characters that are represented

are those which relate to interaction with humans.

Animal characters execute many functions which would be strange for a human character to

perform: unwavering yet wordless loyalty, transport, edibility, an atmosphere of wilderness,

gender and sexual neutrality, cuteness, and a roundabout way to accept the robotic nature of

NPCs. Animal characters offer things that human characters cannot, at least not without

making things weird. In this way, animal NPCs are rushed in where human NPCs fear to tread.

JanA ski offers a thoughtful categorisation of animals in computer games and DeMello offers a

discussion of animal classification systems (46), and this section will not replicate their work.

Instead, let’s take a closer look at the things animal NPCs can do that human NPCs cannot.

[Image removed due to copyright restrictions]

Fig. 1: Concept art for the game Battle Chef Brigade illustrating how each monster can be

processed into different food items. (Arguello)

Animals are companions. They can’t carry as much stuff as Lydia of Skyrim can, but they are

cuter. Other than that, human companions and animal companions do pretty much the same

thing. Companion animals usually don’t have dialogue, which is a relief, but they fight for the

player if there’s fighting to do, and stay by the player’s side. How they affect the player

differently is their status as a pet. Lydia is subservient to the player because she is an

employee, but the subservience of animals comes naturally. Companion animals quickly turn

into virtual pets, a game genre about caring for a nonexistent body. This genre in turn quickly

crosses into the farming genre, where players expect some reward for their care.

In all sorts of game genres, player-characters can ride around on horses, a very wholesome or

archaic form of transport. Given the challenges of horse hair, horse animation, mounting,

dismounting, and player control (Feraday Miller), it’s hard to see why games wouldn’t just give

player-characters a nice flat skateboard instead. Horses make the player-character go faster,

and skateboards could certainly fulfil that function too. But there are a number of other things

63

horses do that skateboards cannot. They give an impression of a pastoral or historical setting.

They are mortal. And combining the player-character’s body with a horse-character’s body

makes the forward motion of the pair an action of the horse rather than an action of the

player-character. They have a directionality to be taken over. The horse is objectified, which

objects cannot be, and this contributes to a player’s experience of power.

Nature, environment, surroundings and animals are easily conflated, both in game literature

as well as game design. Animals often exist in games to make the place look nice and to be

metaphors. The meaning of character becomes blurry in this instance, especially in games

where the animals can’t be killed. The doves of Shadow of the Colossus and The Last Guardian

flee from the player-character, but that is the extent to which they can be affected. In Shadow

of the Colossus, white doves fly around Mono, the game’s damsel in distress. In The Last

Guardian, doves roost in the old castle ruins and distract Trico from his tasks. In the first case,

the doves are a metaphor and exist to add poignancy and a bit of grandeur to the girl’s dire

situation. In the second, the doves are part of the furniture, and exist to make the place look

lived in. In both cases, the doves are an intensifier through what they signify. They add and

concentrate meaning. The player does not interact with them as doves, but through doveness,

the idea of doves.

Isbister writes that when designers don’t want players to relate to characters in a gendered or

sexual way, it is best that the characters be cartoon animals (119). Animality is seen as a way

that players can interact with characters without involving themselves in the messy business

of human social structures, especially the difficult ones like race, gender, and sexuality.

According to Isbister, animals, and especially cartoon animals, are perceived, and relate to

players, as gender-neutral beings (46), and appeal to both male and female gamers (119).

Interacting with animal characters is all of the fun of normal NPC dialogue, without the

complications of human sexuality and gender dynamics, she suggests. The thought of

sidestepping the vagaries of human relations is a pleasant one, but trying to avoid particular

difficulties in interpersonal relations only introduces different difficulties, namely, the

exaggeration of difference between humans and animals.

64

While Berger observes that animals have become associated with childhood, Johansson has

noted that computer games are still seen as media for children (170). Computer games about

animals, then, are contending with two parallel conceits, and must further contend with

dualistic associations. A game about hugging pastel bunnies has different connotations, and

will be attractive to different types of people, than a game about training attack dogs. Even

games that are mechanically very similar, like a game that is about shooting ducks with a

camera and a game about shooting ducks with a gun, will affect players differently. Animals,

then, exist in games to evoke particular feelings that human characters can’t, and they do this

through being seen to possess certain nonhuman qualities and lack certain human qualities.

They are too much in some direction and too little in another. As Berger would say, they’re

further along the spectrum to nothingness, and that makes the player feel more

somethingness.

HEDGEHOGS AND BANDICOOTS

Numerous games, including the Crash Bandicoot series, the Sonic series, A Short Hike and

Night in the Woods, among many others, have a hybrid animal-human player-character. These

player-characters have two things in common: cartoonishness and functional humanity. Their

animal aspects provide narrative justification for personality traits or superhuman abilities, or

their friendly or adversarial orientation towards others. Sonic’s hedgehogness means that he

can hurt others by rushing at them spikily, and that he cares about his fellow animals.

Animality as a costume over a basically human character is exemplified in the case of Crash

Bandicoot: bandicoots are brown, stout, rabbit-like animals who never wear pants, while

Crash Bandicoot is a bipedal, broad-chested, snake-hipped orange humanoid with a fox-like

head. His bandicootery is expressed in his inhuman physical prowess: Crash Bandicoot can

jump high and smash hard. Berger writes of humanoid animals in art, “a person is portrayed

as an animal so as to reveal more clearly an aspect of his or her character. The device was like

putting on a mask, but its function was to un-mask. The animal represents the apogee of the

character trait in question: the lion, absolute courage: the hare, lechery” (18). The hyper-

masculine appearance of Crash appeals to the association of masculinity and animality in

65

terms of wild-ego-driven physical feats. However, the bandicoot is not a well-known animal,

even to Australians. According to one of the designers, the bandicoot was chosen as a player-

character for this very fact (Gavin), as Sega and Looney Tunes, the two major artistic

influences of Crash Bandicoot, were respectively having successes with Sonic the hedgehog

and Taz the Tasmanian devil at the time. Crash’s species, then, does not rely on an association

of bandicoots with a human personality trait, but an association of any esoteric animal with

coolness.

Being a human pretending to be an animal that acts like a human not only occludes the

otherness of animals, but introduces the idea of a universal, bodiless affect. The affect cycle

seems the same, but the humanity of animal-shaped player-characters posits that the beings

engaging in that cycle are interchangeable, which emphasises the cycle as an independent

force. Even when the player is a bandicoot talking to a chunk of wood, the relationship, affects

and personalities of the characters are the same as if the player-character were an elf talking

to a wizard, or two humans talking. Animal-shaped humanly-functioning player-characters

allow the player to play a hegemonic game persona while perceiving it as neutral, beyond any

human specificity.

In some games, the player-character is shaped as an animal, and the player plays concordant

with its appearance. Controlling the animal body is promised as a precursor to inhabiting the

animal mind. Alley Cat, Untitled Goose Game, Shelter and Shelter 2, WolfQuest, Goat Simulator,

and even Frogger offer this opportunity. The bodies of these animals are more a limitation to

movement rather than a total realignment of desires, relations, senses and so forth. Games

with animal protagonists are often an opportunity for the player to roleplay vulnerability. In

these games the player is within the power of others, whether the environment, predators,

human characters or machines. Other than Alleycat, the aim of all of the above-named games is

merely to persist; animal simulation games tend to be survival games. Where animal player-

characters have some modicum of power, like the fish-snatching feline of Alleycat or the sneak-

honking goose of Untitled Goose Game, they are positioned as outsiders, agents of chaos in an

established order. Frogger demonstrates well the tendency of animal player-characters to be

act less like agents than human player-characters. Frogger’s protagonist does not defeat any

enemies, saves any damsels or change any worlds, but rather dodges, avoids, and generally

66

stays out of trouble. The frog is not able to change anything about the situation it is in, and can

only survive by fitting into the rhythm of the system. This reflects a broader idea of the

abilities and positions of animals more generally. There are no enemies or companions in

Frogger, only isolated individuals and the inexorable need to move forward. Frogger’s staccato

movements are not only a testament to the animation limits of the time, but also an indicator

of how little the frog can do in the world. It can move, or it can be still.

While Klevjer has hypothesised that player-characters are a kind of bodily extension – a tool

that the player uses to manipulate the game world – few researchers have studied whether the

purported species of the tool has any effect on the way players interact with nonplayer

characters. Fuchs optimistically proposes that “animal avatars provide potential vehicles for

becoming-animal, thus allowing us to reassess the human condition” (261), though de Castell

maintains that “the kind of world it is within which selves are formed matters as much for self-

building in virtual environments as in real ones – if indeed that distinction even makes much

sense any more” (214), indicating that, for her, the structure of game worlds must be

accounted for in discussing any liberation potential of avatars. While this thesis does not go so

far as to suggest players are just “a cog in a mechanism … subordinated to skilful compliance

to the demands of the role” (de Castell 216), it also cannot conclude the opposite, that animal

avatars can deliver “intimate conjoinings of not only self and machine but also human and

other non-human Others” (Fuchs 264). What it can say, is that in terms of what a player can

actually do in a game as an animal is not all that interesting really, and so the difference in

mechanical affordances of animal avatars and human avatars is either minimal, or makes

animals less capable than humans. Much of the affectiveness of animal avatars, then, comes

from surviving in a game world with few abilities, or ideas about that animal, rather than how

the player engages with others or perceives the environment.

ROBOTS

[Image removed due to copyright restrictions]

67

Fig. 2: The player-character approaches a Strider, a type of robot in Horizon Zero Dawn. (author

screenshot)

That said, animal NPCs are interactive, and do watch and judge the player. Animals in

computer games are, in their small way, animate beings, not mere static images. They can view

the player as well as being viewed. It is possible to be surprised by an animal in a game, to

look across “the narrow abyss of non-comprehension” (5) as Berger extravagantly puts it, to a

creature who operates in mysterious ways from oneself. It is possible to wonder what is

behind the pixels, to question what motivates and instructs its behaviour. For those

inexperienced in computer programming, the answers may be as unforthcoming as those from

a bat. But representations of animals in computer games more often than not contribute to

what Berger calls the “cultural marginalisation” of animals, and represent little more than

repeated ideas about animals and particular species rather than an attempt to recreate reality.

As such, players have different relationships with animals characters than they do with real

animals, though as Lin et al. notes, they are constantly comparing them. The relationship

between players and animal characters in games is unique, incomparable to players’

relationship to humanoid characters or to real animals.

Horizon Zero Dawn is a AAA game about running around the countryside and hitting robot

dinosaurs with a huge sword (as well as uncovering the mystery of the AI apocalypse that

resulted in the collapse of human civilisation). Its protagonist, Aloy, is a down-to-earth, no-

nonsense, rough-and-tumble girl of extraordinary strength and millinery prowess. On first

glance, she lives in a world where robots have replaced animals entirely, yet this radical

ecology seems to be taken for granted by the hunter-gatherers around her. The robots are

representations of representations of real animals: they look like robot versions of familiar

animals. In the game ecology, they perform the same roles as animals would, and in the game’s

mechanics they also perform the same. Robots seem to take the place that animals would in an

imagined primitive people, acting as totems, resources, mythological figures and the centre of

the economy. Robots take the place of animals not just in terms of human exploitation, but

within the world too: eating grass, trampling the ground, reproducing and performing sundry

ecological services. There is another important similarity between these robots and animals –

they engender the question: who made them and who put them there? The search for this

68

existential answer ends up being the main quest of the game, and has an easy and

uncontentious solution. Turns out, after an evil AI destroyed all life on earth, a good AI tried to

terraform the remains with these robot animals, eventually planning to replace them all with

real animals, and restore human civilisation, when conditions were right. But the plans never

came to pass and the planet ended up with a large number of robots, a small number of

animals, and a subsistence society.

Aloy’s main task in the game is to ‘kill’ lots of robots. Different robots require different fighting

tactics; the player must learn numerous strategies of approaching, attacking, tricking, or

hiding for each of the species. Familiarity with the capabilities of a variety of animals is

essential to survival in the game. The mechanical animals, however, comprise 25 species and

exhibit a variety of behaviours. In concordance with the game’s narrative of ecosystem

collapse and technological deliverance, and presumably for the pleasures of deathly

accomplishment, they are all in the guise of megafauna: bison, buffalo, horses, panthers,

dinosaurs, and terrorbirds. The robots each have roles to play in both the narrative and the

game’s ecosystem. The bison and buffalo, for example, consume grass and turn over earth,

performing the same actions that a biological ruminant would (although it is unclear what

happens to the grass after consumption). Predators watch over the herbivore herds, alerting

the beasts with alarm calls and mobilising into attack formations when a human is sighted.

The player kills the robots for a number of reasons: to procure metal shards (a type of

currency), to acquire machine parts for constructing new weapons, to fulfil a quest (such as

acquiring machine parts or removing a threat to a village), to access a new area, to show

prowess to human NPCs, for fun, or for practice.

The player has a number of weapons that can kill the mechanical animals. One of these

weapons can reprogram the robots to act as the player’s ally. It is difficult to wield, as the

player must slowly sneak up on a robot without them becoming aware of the player’s

presence, and then must stay perfectly still and undetected for a number of seconds. It is often

easier just to kill a robot instead of trying to reprogram it into a friend. But once achieved, this

reprogramming results in a remarkable reversal of the robot’s behaviour. Once forcibly

converted into a friend, instead of attacking humans and defending robots, the allied subject

defends humans and attacks other robots. It is also able to be mounted and rode as an ersatz

69

horse. Its loyalties are completely inverted; what once was loved is now hated, and vice versa.

This can have some difficult moral consequences. Once reprogrammed, a robot is destined to

live alone. Reprogramming it results in a life without companions. Some quests require the

player to kill all robot dinosaurs in an area to progress, including any reprogrammed allies.

Faced with this choice, players may choose (as this researcher did) to return to an earlier save

and fight the battle again with no robot allies, rather than kill a friend (even though reloading

the game effectively wiped the friend from existence, albeit in a way that displayed no violent

animations and was external to the game narrative).

The ability to reprogram robots draws attention to the fact that they are programmed in the

first place. They are not creatures of judgement or independent thought, they are vessels of

cause and effect. And in this, they are not differentiated from the game’s animal characters.

Relationships with human NPCs in HZD can change based on dialogue choices and actions. The

player can have many different relationships with the same human characters depending on

what they say and do. But no matter how they act towards animals, the robots and animals

will always relate to the player one way, unless forcibly reprogrammed. This also makes

reprogramming less an act of psychological violence and more like changing the settings on a

microwave. The creatures never had free will to begin with, so changing their relationship to

the world is not a fraught act. It is not presented in the game as a morally dubious thing to do

(neither is killing them), just another cool ability.

As well as becoming friends, any robot can also be “corrupted”. Corrupted robots are stronger

and more aggressive, and live in corruption zones – especially difficult areas of the game.

These are essentially more extreme versions of ordinary robots – faster, more perceptive,

more violent, more passionately intolerant of human presence. They promise an intensity of

what the player has already experienced. They cannot be turned into friends. But this solidity

is also positioned as an aberration. They are “corrupted”, that is, they have been affected, they

have deviated from the norm, they are other than their natural baseline state. When Horizon

Zero Dawn posits that its robots have a basic state which only external actors can alter, it

reinforces the synonymy of normal and natural, and suggests that game animals are passive

beings that can only be acted upon. No robot in Horizon Zero Dawn decides to be a friend to

humans, nor to be super-aggressive. They are only programmed by others to be so. The

70

wilderness and its creatures stays in stasis until something acts upon it. In this, Horizon Zero

Dawn reveals that its massive robot dinosaurs are robots pretending to be robots pretending

to be dinosaurs, but they were actually just dinosaurs all along.

The robots in Horizon Zero Dawn behave functionally like animals in games: that is, they

behave like robots. Most games just have actual robots pretending to be animals but acting like

robots. Horizon Zero Dawn acknowledges and celebrates the artificiality of its creatures. The

robots in Horizon Zero Dawn behave like any other animal represented in games.

The artificiality of Horizon Zero Dawn’s robots is emphasised in a number of subtle but

repeated ways. Firstly, the player, through a special device, can see the paths the robots will

walk in the future. Secondly, if interrupted, the lights of the robots switch from blue to yellow

(cautious curiosity) or red (active hostility). This visible display of affect is designed for the

player’s ease of sneaking, but this obvious show of intent implies that robot emotions are

discrete, linear reactions prompted by external events and internal timers, easily

distinguishable, and work in quite different ways to how emotions function in living creatures.

Thirdly, their appearance is metallic and hard, comprised of panels, tubes, wires, blocks and

tanks of green liquid, and their voices are machine-like whirrs and bleeps. Fourthly, the robots

are uniform, with no baby robots, thin robots, fat robots, beautiful robots or elderly robots in

the population. Fifth, the player can destroy them in a modular way, blasting components off

them one by one. Finally, within the human population, there is no reverence for the robots,

only trepidation and technical knowledge. The majority of humans do not mourn or celebrate

the souls of the robots, and the one human that does attach some mythology to the individual

robots is depicted as off his rocker. In this way, the artificiality of the robots serves both to

maintain the dualism between natural and artificial. The mindless, clockwork robots that so

efficiently take the place of animals affirm nature as a passive environment for conscious

agents to do their work.

In the beginning of the game, the player is encouraged to relate to the robots the same way

they would relate to any digital animal: to become familiar with their mechanical behaviours,

develop methods for interaction with them, and to extract resources from their defeated

corpses. Having discovered that the robots are normal game animals, the player can relax. The

71

player knows where they stand – at the top. But because humans have few real-world

obligations to giant robot dinosaurs, HZD offers some unique affordances to particular game

audiences. Horizon Zero Dawn is a game that finally lets vegetarians participate in violence

against digital animals without guilt (McCormick). The game relies on players caring about the

game ecology enough to keep playing and slowly uncover its history, but not caring about the

ecology enough to relate to the robots in the same way as any other hunting game.

Coghlan and Sparrow’s discussion of the ethics of killing digital animals in computer games

touches on the usual arguments about killing things in games (it makes people violent,

immoral and prejudiced), but with an interesting addendum. They concede that the worst that

the representation of animal cruelty in computer games can do is have “at least some

potential, even if only a modest one, to contribute to moral indifference toward animals” but

argue that it’s bad for a different reason: players don’t want to do it. They observe that many

players do everything they can to avoid killing virtual animals, to the point that their game

progress suffers for it. Chittaro and Sioni’s work investigating the desensitising effects of an

ant-themed game of whack-a-mole deflates this conclusion a little. They found that not only

does squashing ants not make players desensitised to violence against animals, but removing

the ants and replacing them with squashable rectangles makes the game boring.

Representations of ant death in their game was important to player satisfaction, but was

either not important to the formation of attitudes towards ants or reflects existing attitudes

towards ants (236).

Instead of viewing player discomfort at killing digital animals as evidence that it is a bad thing

to do, Stanley sees it as evidence that digital animals are affective. Feeling upset about the

death of animal NPCs, whether as a player or as a researcher, is “another fantastic example of

how the medium is striding towards more meaningful measures of interaction … it’s a good

sign of a maturing industry when we can witness a dragon hunt of all things and feel a genuine

emotional response” he argues. Johnson and Verdicchio also offer a neat explanation about

why it’s actually fine to kill as many computer game animals as you like: they don’t care

whether they live or die. While analogies are often made between robots and animals (293),

these analogies are always lacking, Johnson and Verdicchio argue, because they always discuss

animals and robots entirely in terms of their relations to humans. That the way humans

72

perceive and act towards robots and animals might be similar in some cases, they emphasise,

doesn’t change the fact that the two are actually very different.

A few authors argue that hunting games do not have to have lasting moral effects on players to

be problematic; they can just be bad games. Sawers and Demetrious’ rather fantastic paper on

the political implications of American hunting arcade games in Australian quasi-public spaces

characterises hunting games as “a space in which the effects of risk society [a society

concerned with the risks of the future and how to deal with them] are temporarily forgotten.

The domination of nature … can be played out in the game, without the associated risks”

(244). Hunting games, in their view, are a beautiful escapism from an environmentally

degraded world into the kind of world that caused all our problems in the first place. However,

they are also “forms of political persuasion working through a new modality” (244). van

Ooijen suggests that even though animals and robots are different, the ideological functions of

game mechanics represent a certain attitude towards animals, which the player has a hard

time escaping (33). He argues that “that games make meaning, not by creating a full

reproduction of the world, but by selectively modelling certain procedures associated with

specific phenomena” (35), and that the specific selections “constitute tangible models of

specific areas of ideology” (35). That some represented bodies in games are able to be killed

and eaten and some are not places these beings in a hierarchy defined by regulation. Like

Stanley, van Ooijen sees the indignity of animal NPC death not as a glorification of killing but

as a realistic representation of real animal deaths. The mechanics of looting, though, show the

fundamental position of human and animals: “When you loot a human body, you acquire their

possessions: weapon, armour, money, and so on; but when you loot an animal body, you rather

acquire its meat. At its most basic structural level, then, the game presents the argument that

humans are robbed, whereas animals are hunted” (38). The mechanics of respawning, too,

“model natural life in accordance with the ideology of industrial capitalism, i.e. as an infinite

source for endless exploitation.” (41). van Ooijen observes, though, that a number of games

have tried to tackle this thoughtfully, albeit clunkily (37-41), mostly by limiting the kinds of

animals that can be killed, or having the number of type of animals killed affect a morality

mechanic. HZD allows players to sidestep all of this by representing their characters as robots.

They afford an instrumentalised relationship to animal NPCs that isn’t made fraught by their

73

representation as real animals. By representing their robot animals as robot animals, they

invite players to see them as merely robot animals.

Plumwood observes that the mechanic of hyperseparation is a key part of creating otherness.

HZD’s robot dinosaurs clearly stand out from the game’s lush, picturesque grasslands. Visually,

they provide a striking contrast. They are shown as having an essential nature completely

different from all the other beings in the game. Mechanically though, they’re not any different

from any other animal NPC. Plumwood notes that one popular way of maintaining

hyperseparation is the distinction of things as sacred and profane (49). The robot dinosaurs,

although huge and powerful, are not objects of worship for HZD’s humans. Instead, like other

animals, they are resources, in a class of their own. When Aloy is asked by a villager to kill a

robot in order to save their town, the game is asking the player to see the robots as profane, a

threat to something which must be protected and isolated (Plumwood 50). The robot

dinosaurs, being huge and artificial, demonstrate an “unbridgeable separation, a separation

not open to change” (51).

PIÑATAS

The 2006 game Viva Piñata offers a less destructive vision of artificial ecology. Viva Piñata is a

game about collecting cute creatures, and as such finds itself within the simulation genre

alongside farming and zoo management games. As in Horizon Zero Dawn, Viva Piñata

introduces the player to a degraded environment that only human labour can restore. The

primary task of the player is gardening: tilling soil, weeding out unwanted plants, sewing

seeds, watering seedlings, and deciding where plants, ponds and animals should go. Colourful

paper animals show up and live in the garden once a minimum amount of work has been

done. Different animals have different needs, and so the player can decide to lure a particular

animal by designing particular habitats for them, growing particular plants, changing

particular landscape features. The Buzzlegum, for example, will only appear if there are two

buttercups and no Arocknids in the player’s garden. Unlike Horizon Zero Dawn where the

humans and robots have an uneasy relationship of unassailable difference, the pinX atas and

74

first-person player of Viva Piñata are in a mutual relationship of cause and effect, becoming

more complex over time. While the human has god-like oversight of the system and the

pinX atas are ultimately dependent on the player to survive, their needs are at the centre of the

game, and the player is focused on their individual welfare and preferences. The player’s goal

of wilderness rehabilitation is gamified by the collection of colourful creatures.

Though pinX atas have a reputation for being extremely smashable, the pinX atas of Viva Piñata

cannot be damaged physically, only emotionally. The player must keep their habitat well-

maintained, otherwise the creatures will become sad and leave the garden, and the player will

find themselves owning an incomplete collection. Management of the animals becomes a

concern when there are enough of them to breed, and predatory pinX atas start to arrive, but

otherwise the pinX atas find their own food and entertainment. The main pleasures of the

pinX atas are their colourful appearances, the pride of being a competent enough gardener to

attract them, and the satisfaction of completing a set. In this way, the appearance of different

species of pinX atas serves as an indicator of progress – pinX atas arrive in the garden only once

some work has been done, and as the game goes on, the player has to do more work, and more

difficult kinds of work, to attract more pinX atas. The greater quantity and quality of pinX atas the

garden contains, the more successful the player is. The pinX atas are both the game and a

scoring system. Unsurprisingly, as adorable and characterful as the pinX atas are, this has

implications for the relationship between the players and the animals.

The relationship the player has to the pinX atas of Viva Piñata is caring and understanding. The

player gives the pinX atas what they need to survive, and in return receives the continued

survival of pinX atas. While not quite a farmer, the pinX atas are a livestock of a kind, a herd that

produces not meat nor milk, but a sense of achievement. Unlike the animals of farming games

like Stardew Valley, they do not produce anything and do not perform labour for the player.

They are not commodities in the sense that they can be exchanged, but they are not quite pets.

Interaction with them occurs on the level of species, not the level of individuality. When

problems occur, they must be fixed by the player, not solved democratically or left for the

animals to figure out alone. The pinX atas have some capacity for self-determination, as

evidenced by their ability to survive in the wild, their decision to come into the garden or leave

it if it’s not up to standard, and their ability to feed and entertain themselves once ensconced

75

in their allotments, but they need the player to make a place for them to live, to keep them safe

from predators, and to build an ecology. They need, in short, the player as a benevolent

authority; a zookeeper.

Zoos trade in exoticism, as DeMello observes (101), and they also serve as a dissatisfying

container for the otherness of animals, Berger remarks (21). They promise a meeting with

animals but stand as a frustrating “monument to the impossibility of such encounters” (Berger

21). “Adults take children to the zoo to show them the originals of their ‘reproductions’” (21),

Berger argues, but Viva Piñata does away with the pretence. Animals have always been a

demonstration of technological spectacle, Burt argues, and “changing configurations of

visibility and invisibility … are what determine both the nature and power of animal

representation (Burt 222).Viva Piñata shows that the display of animals to demonstrate

technological achievement has not vanished, just modulated. That the player does not act on

the creatures, but can only modify their environment, reveals the network of affective

relations they are embedded in – or more accurately, individuated from. PinX atas come into

existence when circumstances are optimal, and while the game’s narrative posits a wilderness

that the pinX atas come from and go to, mechanically the pinX atas are rendered into view upon

the completion of certain player-instigated conditions. They are entirely products of their

environment, and it is up to the player to manage that environment. The game of Viva Piñata,

while ostensibly about a little zoo of pinX atas, is actually about the modification of an affective

landscape, such that certain things and not others are individuated into existence.

DOGS

As Burgess-Jackson notes, it is unproductive to refer to animals and relationships as if they

were abstract, undifferentiated blobs (159), which is perhaps the point at which it’s time to

leave Plumwood’s structural-based approach and move to Haraway’s concept of significant

otherness. Haraway’s work details the philosophical implications of close relationships with

domesticated nonhumans. She remarks that “Just who is at home [in the mind of the dog] must

be permanently in question. The recognition that one cannot know the other or the self, but

must ask … who and what are emerging in relationship, is the key … We [humans and dogs]

76

are not one, and being depends on getting on together” (50, emphasis in original). To be in the

world requires getting along with things that are both completely different from oneself and

are necessary for constituting oneself. This is true of all animals, but comes into sharp focus

with dogs, because they have such intimate relationships with humans.

Kirk argues that people like dogs more than other animals because they have a greater sense

of ownership and control over them. Kiesler et al. further propose that a sense of ownership

over an animal means that human owners perceive more complex psychological explanations

for their pet’s behaviour. Lin et al. also note that the intensity of emotional expression and

player control of digital dogs (which they inexplicably do not refer to as dogitals) account for

much of their appeal to players of pet games. Chesney and Lawson, however, observe that

players of Nintendogs do not enjoy or feel as attached to their virtual pets as much as owners

of real dogs enjoy and feel attached to theirs (though Consalvo and Begy show that the death

of virtual pets can be quite affective (95)). To further emphasise the difference between the

affectiveness of pets and the affectiveness of virtual pets, Warpefelt and Verhagen

uncharitably define pets in games as “essentially a purely decorative NPC that … is simply

there as a vanity item” (7). Like friends, the virtual pet is made to replicate familiar

relationships, but in simplifying the relationship for digital spaces, it reveals quite a lot about

what can be considered the fundamentals.

Pet ownership is a common human-animal relationship, and also a relationship that game

designers are keen to represent. It is therefore unfortunately necessary to look at the real

world for a second to see what designers are trying to aim for. Affection is at the centre of the

human-pet relationship, though not unproblematically, as Haraway notes (33). Archer has

shown that for many people, the death of a loved pet is comparable to the death of a loved

human (239). As Archer notes, humans can be extremely attached to their pets, even to the

point of spending money and time on them (238). While pet ownership involves significant

caregiving and nurturing, it is also the case that humans find strength and security in their

companionate relationship, and rely on their animals for their emotional needs (Archer 245).

As such, humans acquire pets for the express purpose of affecting and being affected by them.

Burgess-Jackson notes that pet-ownership is inherently other-oriented: “We are attuned to

their material, psychic, and social needs” (160). This affectability is helped, in the case of dogs

77

and cats, by the perceived similarity of their emotional lives (Archer 251). Archer argues that

“the emotional similarity that people recognize in animals … forms the basis of being able to

communicate with them by visual and auditory signals, and by touch (the equivalent of

grooming), and to share object play with them” (251). Whether animals actually do feel the

same emotions as humans is moot, as it is the perception of emotional similarity that is

fulfilling to the pet owner. Archer, in accordance with Darling, Isbister, Reeve and Nass, and

Nass and Moon, asserts “the human tendency to project feelings and thoughts onto animals,

and even onto inanimate objects such as robots and computers, would seem to be a pervasive

one … Anything that is similar to a human being, and with which a person has repeated

interactions, is treated as if it has a ‘mind’” (252). This is great news for game designers, as it

means that replicating human-animal relationships in games should be easy.

In the recent virtual pet game Wobbledogs, god-like players feed, clean, praise, scold, train, and

manipulate the microbiome of wobbly oblong dogs. Lin et al. (4) and Consalvo and Begy (99)

observe that the main audience for virtual pet games is tweens, and that the vast majority of

virtual pets are eventually abandoned out of boredom, but Wobbledogs attempts to avoid this

fate by giving dogs the abilities to age and die, and players the ability to breed and mutate

their dogs and decorate their dog habitats. The player of Wobbledogs does not only have a

relationship to the individual dog, but a responsibility to their whole species, as they manage

mutations and selectively breed dogs to achieve weird effects. The modification of the virtual

dog body does not begin and end with virtual dog biscuits, but in their genetics, limbs, and gut

flora. In her discussion of Spore, Chang notes that such mutation mechanics promise “an

ethically unencumbered space in which players can spool out countless environmental

futures” (33), but they also firmly place the human as master, an “architect of fantastic worlds

rather than an individual within them” (Chang 31). Selective breeding can make wobbledogs

so tall that they cannot reach down to their food bowl and must be hand fed: the player’s

decisions affects how the wobbledog experiences its world. The player has greater control

over their dog’s body, which becomes an instance of a larger whole rather than an individual.

Molleindustria’s Dogness takes this philosophy to the extreme in its dogpark simulation about

“the resurgence and normalization of white supremacy in the Trump era” (paolo 2018). In it, a

god-like player must breed dogs to create the perfect pup, but is hounded by natural variation

78

and the threat of mongrels. These dogs do not need food, water, or shelter, they need pure

genetics and strong management. paolo notes the strong connections between purebred

animals and human eugenics, and this game, like most of Molleindustria’s games, observes

that game mechanics perform ideological work. The affect of what games signal through

metaphor is as important as what they superficially present as, and that animals are

commonly used in games as metaphors for humans means that animal games carry a lot of

weight.

That said, in SokPop’s Pupper Park, the player plays as a puppy running around a dog park.

This cute game is about the unproblematic experience of joy, as players run, bark, pick up balls

and sticks, and drop them again. The player is given little instruction in Pupper Park, and must

work out the controls for themselves. This speaks to what Fuchs, in a discussion of Bear

Simulator, calls the becoming-animal, a recognition that “before humans can even start to

ponder philosophical or religious questions, they must learn to control their bodies, to move,

and to navigate the world, so do bears … and so do players, as they begin to interact with—

and act through— their bear avatars” (266). The pups of Pupper Park will never start to

ponder philosophical questions, and that is their whole appeal. They promise a world in which

theses don’t matter, work never has to be done, and getting the most fun out of life is the only

consideration. The player can inhabit the imagined joy of the carefree puppy as “the video

game opens up a new state of being, as players are thrown into a world that is not their own

and a subject position not their own, but which become their own through gameplay” (Fuchs

267). However, even the game developers note that the game is an entertaining diversion of

ten minutes or so. Pupper Park is a game made for moments of in-betweenness, to forget the

boring human world and imagine something better, but only for a short time between other

things.

John Berger was definitely a cat person. Of keeping pets, he says, “in this relationship the

autonomy of both parties has been lost (the owner has become the-special-man-he-is-only-to-

his-pet, and the animal has become dependent on its owner for every physical need), the

parallelism of their separate lives has been destroyed” (15). The genetic manipulation of

generations of virtual dogs points to their objectification, as “they are conceived of as [the

master’s] instruments, a means to his ends … he is free to impose his own ends” (53), as

79

Plumwood says. “The identity of the underside [in contrast to the master’s upperside] is

constructed instrumentally” (53), and in the charming Wobbledogs and the self-aware

Dogness, this is quite literal, as dogs are constructed by the player for their own ends. But

Haraway sees the malleability of dogs as an opportunity, denying that the parallelism of dogs

and humans has been destroyed, but rather can come to an uneasy balance. The power

dynamics of dogs and humans, she suggests, are commonly forced into master-other relations

by people who use the word alpha in all seriousness (48), but are more productively managed

by ideas of companionship. To have a significant other requires becoming a mixed breed, she

suggests, and recognising that “the world is a knot in motion” (6) and that everything

seemingly fixed is actually constantly in contention.

CHICKENS

Much work in recent years has criticised Huizinga’s theory of the magic circle – the idea that

game space constitutes a realm entirely separate to the real world. Instead, a new theory has

developed that digital spaces, and game spaces, are an extension of, auxiliary to, or an equal to,

real spaces. As a consequence, interaction with others in digital spaces is seen to have some

things in common with others in other spaces – even if the rules and norms are not the same,

there are still rules and norms. Games are not a completely separate place, but they are a

particular place. Player interaction with digital animals highlights both the otherness of game

spaces, and the ways in which game spaces both reflect and act upon real spaces.

For most people reading this, animals such as deer, horses, cows, rabbits, pigs, goats and

chickens are a familiar idea but not a familiar presence (alive, anyway). But for game players,

digital livestock is ubiquitous. Interestingly, even though farm animals are alien to everyday

life and interactions with them in real spaces are rare, interactions with them in digital spaces

are not particularly fraught, or even significantly novel. Although these creatures are rare and

strange, and players have probably not had the benefit of extended interactions with them in

real life, interaction with farm animals in game spaces is easy, even to the point of being

humdrum. Easiness is often a result of repeated encounters, something that has been learnt

previously. In the case of the chicken, although the reality of the chicken is alien, the idea of the

80

chicken is an old friend. To take it further, although the real body of the chicken is alien, the

digital body of the chicken is an embodiment of the idea of the chicken. And players have often

spent more time with the idea of the chicken through media representations of chickens, than

they have with non-represented chickens. Chickens in computer games are different to real

chickens, toy chickens, filmed chickens or literary chickens. People expect different things

from game chickens, and interact with them in different ways. One of the more common ways

that players interact with digital chickens is kicking. One might not kick a toy chicken because

it is the property of another who you do not wish to offend, or because is an object of some

value. One might not kick a real chicken because it belongs to another, or because one has

some sense of compassion for living beings. One cannot kick a chicken in a movie or book at

all. But one may kick a digital chicken without consideration for ownership, sentience, or even

damage. One might even get rewarded for the kick in terms of spectacle or even game points.

Kicking a digital chicken in a game is loads of fun – they usually make a scene of fluttering and

squawking, and return back to benign normality quickly. The immediate cause and effect is

very satisfying. It demonstrates to the player their power of activity, the extent of their

affective capacity. It’s very tempting to abuse the chicken-shaped computer program that has

been specifically designed to be abused for fun, even for those of us who are normally quite

nice. Animals in games are an interesting case study in character objectification specifically

because they are seen to have less agency and to be somewhat automatic in the real world.

Their presence is evocative and inviting in a way that other beings’ existence is not.

This is not to say that there is some pure chicken outside of human culture that can be grasped

if only media representation of the animals were more accurate. But there is a chicken depth

that can be plumbed to a greater or lesser extent, and a chicken thickness that can be

appreciated. Fothergill and Flick discuss representations of chickens in computer games and

connect their in-game roles to the way humans have historically used the birds, and how they

are currently viewed. Through this, they show how changing social attitudes towards our tasty

friends have influenced depictions of chickens in computer games, and the kinds of

relationships players are expected to have with them. Chickens are not an uncommon sight in

games: in medieval-esque settings they can often be found roaming in villages, in other games

they are used as food or props. Fothergill and Flick argue that the digital chicken is an

extension of the domestication of the chicken, a furthering of the conversion of the wild

81

animal into something interdependent with humans. In their digitisation, certain qualities of

the chicken have been enhanced, and others pruned, and the selection of these qualities is

indicative of broader social ideas about chickens. Invariably, the qualities of the chickens that

have been enhanced are those borne of the age in which chickens are primarily known as food

sources: chickens as silly, fearful, unintelligent creatures. The authors note: “The way in which

humans perceive other animals impacts how we treat and raise them, which in turn leads to

physical changes in some animals (e.g. increased size, rapid growth), which then perpetuate

and deepen our views of that species … A lack of understanding of past chicken-human

relationships can lead to portrayals in video games which may only serve to normalise

negative ideas about the real animal” (102). In this, they argue that domestication, and

specifically digitisation, creates ideas about chickens that are only further reinforced. They are

highly mediated before they even appear on the screen.

According to Fothergill and Flick, chickens have been used for religious and entertainment

practices much longer than they have been considered a source of food (102). These uses

continue in games, with a number of titles depicting chicken-shaped gods, or eggs as symbols

of renewal. Although it is difficult, as the authors note, “to detect whether deific, supernatural,

and sacred manifestations of chickens in video games are somehow connected to this ancient

association or are presented in these ways as an ironic joke by designers who sought out what

they perceived to be an unremarkable creature” (106). Chickens may be portrayed as gods in

some games precisely because of the incongruous combination of godly omnipotence and

bestial inefficacy. The same can be said for chickens as protagonists (104), as well as the

opportunity to make countless jokes about cocks.

Fothergill and Flick comment “One of the most prevalent ways that video game players

interact with chickens is through violence” (102), and that in “in Far Cry 4, Grand Theft Auto V,

Crysis and Counterstrike: Global Offensive, chickens exist to be shot at and killed, with no

particular reward” (103). Chickens are represented in games if not for the explicit purpose of

players kicking them down the street, then in the knowledge that this will be a consequence of

their inclusion. Fothergill and Flick hypothesise that violence towards chickens in computer

games is merely a continuation of ancient entertainments such as cockfighting and

cockthrowing (stoning chickens to death for fun) (105). They argue that player-chicken

82

violence is a consequence of the “hyper-masculine” nature of games, in which the domination

of nature is a key part (104). Arguing that computer games have replaced the outdoors as a

“male play space” (104), the authors observe that “masculinity is a complex concept, and

video games tend to fall into the trap of portraying masculinity (and male characters) as part

of the hegemonic masculinity of macho, domineering, rigidly ‘manly’ men” (104). In this

culture it is apparently inevitable that chickens will be persecuted.

But violence towards the digital chook does not always go one way: in some games, chickens

fight back. However, according to the authors, the existence of the chickens’ will to self-

defence is just another confirmation of their reputation as “cowardly, unimportant,

disposable” beings (106): “Chickens in games which are very powerful or respond in kind to

violence are … intended to be unexpected, which further entrenches the conceptualization of

the chicken as a simple object which a player can attack ‘to see what happens’ or because it is

perceived as humorous” (106). It is the innocuous reputation of the chicken (in stark contrast

to their history as fighters) that make chicken self-defence so engaging.

Player-on-chicken violence is not an inevitable characteristic of every game in which chickens

feature – unless you think of modern agriculture as systematically violent. Plenty of games

also exist where chickens are farmed, tended, herded into pens, bred, or kept as pets. It could

be said that farming games also represent domination over nature, and that farmed chickens

aren’t so different from kicked chickens in their provision of player affect. The kicked chicken

gives a funny experience, and the farmed chicken gives an egg. Farm games, although pastoral

and wholesome, perpetuate the idea that a human can do whatever they want with an animal,

even if what they want is the animal’s happy welfare. Chang notes that “While we might decry

[the hyper-industriousness of virtual farms] as evidence of our human tendency to impose

analytical frameworks and capitalist deliverables on a qualitatively rich natural world, as a

player these minutely measured levels of readiness are reassuring, because they render nature

predictable and therefore manageable” (43), that is, pleasurable representations of nature can

lead players to think more favourably towards non-represented nature, in what Fothergill and

Flick call “cycles of perception” (102). She identifies farm games as part of a pastoral literary

tradition, and suggests that the “garden in the machine” (47) both harks back to the romantic

ideal of family farming and obscures modern agricultural working conditions.

83

[Image removed due to copyright restrictions]

Fig. 3: A zoomed-in view of a chicken coop, four chickens, two ducks, and one player-character in

Stardew Valley. (author screenshot)

Stardew Valley is an addictive indie game about growing thousands of tonnes of food and

never eating it. In Stardew Valley, players play as a subsistence farmer starting from scratch,

and must plant, water, and harvest crops, as well as maintain a basic social life with the nearby

town’s inhabitants, mine minerals, and marry Elliot. Chickens are the cheapest of the available

livestock, and so are usually the first the player will own. They consume hay and make eggs of

varying qualities. These eggs can be sold, or they can be made into mayonnaise and sold at a

higher price. Though the player-character can eat food, there usually isn’t any need to, and so

all foodstuffs the player produces can be sold for money, which can be used to buy more farm

equipment, which can be used to make more and different foods, which can be sold for more

money, and so on. The player is on the perpetual edge of being able to buy something new, and

this, not hunger, is the reason for their food production. As one of the first animals, chickens

are the backbone of this production.

Stardew Valley’s player interacts with their chickens via three objects and six actions: a coop,

hay, and happiness; and initial purchasing, coop building, hay allocation, pats, door opening,

and egg collection. A typical interaction with chickens would go something like: asking the

builder Robin to build a chicken coop on a cleared plot of land having saved up enough

materials and money, purchasing chickens from the rancher Marnie (which also involves

choosing their names), and then, each day, opening the chicken coop’s door, patting each of the

chickens, checking the hay levels, collecting the eggs, and closing the coop door at the end of

the day. On sunny days the chickens go outside and wander around, and on rainy or snowy

days they stay inside their coop. They lay eggs at night. They are either white or brown, so

once the player has more than two it can be hard to keep track of them all (though the latest

version of the game introduces blue and gold chickens as rewards for certain quests, and

mods can change the chickens into basically any colour). Chickens must be happy in order to

lay eggs, and they become happy by being patted. They want to be acknowledged by the

player, if only for a second. But the mechanics of chicken ownership in Stardew Valley is

84

unsurprisingly more like tending a machine than a pet, with more actions dedicated to

maintenance activities around the bird than the bird itself. Chang argues that in this top-down

manipulation of workers, farm games are more like military simulations or real-time strategy

games in that they “model the extractive logics of resource use and development. Both stress

resource gathering, structure building, and the manipulation of agential units” (Chang 67).

Chickens, then, are the troops of this pastoral army, fighting to give the player everything they

need to feel powerful.

That Stardew Valley’s chickens are indistinguishable from each other points to another of

Plumwood’s othering mechanisms, stereotyping, or homogenisation. It’s a pretty simple one:

“differences among the inferiorised group are disregarded” (53). “To the master” she says,

“residing at what he takes to be the centre, differences among those of lesser status at the

periphery are of little interest or importance … to the master, all the rest are just that: ‘the

rest’, the Others, the background to his achievements and the resources for his needs” (54).

The chicken in Stardew Valley “is related to as a universal rather than a particular, as a

member of a class of interchangeable items which can be used as resources to satisfy the

master’s needs” (54). Furthermore, they have an essential nature that is fixed and makes them

easily identifiable and predictable. DeMello notes that “humans so easily draw on animals to

make sense of human realities” (284), but it works the other way too, as Berger observes, “the

pettiness of current social practices is universalised by being projected on to the animal

kingdom” (15 emphasis in original). Animals, made into blank slates, reflect humans back to

them; the sameness of the chickens underscores the specialness of the human.

HORSES

If one wants to move at a speed faster than a jog through the Tolkienesque fantasy world in

which so many games take place, one must use some sort of genre-appropriate transportation.

And that transportation, more often than not, is a horse, ridden bareback. Usually a relatable

brown horse, though sometimes a dirty white horse, or a black horse if your player-character

is particularly bad-ass. But no matter the colour, it is rare to find a dark horse in video games.

Instead, horses are completely (motivationally) transparent. A horse that comes close to

85

opacity is Agro of Shadow of the Colossus, who faithfully carries our hero from tragedy to

tragedy, and yes, sacrifices herself in an act of heroism to save the player-character in a

dramatic climax. Agro reacts emotionally to game events, shows independence, displays

idiosyncrasies, and is affectionate towards the player-character. It is not the complexity of

interaction that engenders affection for Agro, nor equality of position. Agro is at the player’s

beck and call, and obeys any instruction the player gives her. But Agro has her “own

personality and presence in [his] own right, rather than [as a] mere gameplay accessory to the

player” (T. Cole 3). Like many digital horses, Agro is the only friendly presence in an otherwise

hostile world. Horses are non-judgemental witnesses and travel with the player to, or help

them escape from, memorable situations. Cole observes that it is “Agro alone that the player

can communicate with, interact with in a positive way and work with to progress through

[the] game and this extra focus on the relationship, without competition from anything or

anyone else, encourages a closer bond between player and partner” (4). Although Agro is a

horse, a type of creature with which the player probably does not have a deep familiarity, her

qualities of exclusive co-presence, co-operation and communication make her a close friend.

A horse helps you arrive. There is no comparison between galloping into town on a steed and

shambling through the gate on foot. An entry on horseback lets the player feel grand and

confident. It takes time to learn how to control a horse, so riding a horse well is an indicator of

experience and mastery. Sometimes, horses cost money, so owning one can also be a sign of

wealth. A horse also confers a commanding height upon the rider, but this doesn’t translate to

the third-person perspective, in which the camera is usually placed above and behind the

player character anyway, so the player looks down on the world as a matter of course. It is

mostly a metaphorical level to which the player character rises, when they join forces with

horses.

A horse helps you escape. In Skyrim, horses are faster than most game enemies, so it is

possible to outrun enemies without killing them. In The Legend of Zelda: Ocarina of Time, the

player can easily ignore all enemies (except poes) by galloping away from them. In Stardew

Valley, a horse makes the player move faster, saving precious time that could be spent later on

watering extra crops. In Red Dead Redemption 2, the player can kill a rider, steal his horse, and

86

use it to carry the corpse back into town. In Shadow of the Colossus, getting lost in the

countryside with Agro is a relaxing reprieve from the intensity of colossus battles.

Horses are also awkward to mount, awkward to control, and awkward to dismount.

Interacting with a horse is where game animation is tested to its limit. Player-characters do

not naturally connect with others, outside of a cutscene. Players can find their characters

dismounting the horse accidentally, mounting and dismounting in quick succession, or

galloping on the spot while caught in a bush. Huang et al. note that animating horses is “a

challenging problem as a horse has six different gaits and changes its gaits at different speeds”,

which is difficult to transition seamlessly. This awkwardness, though, has a beneficial effect –

it makes it clear that the horse and the player-character are two separate beings. Although

they are conjoined once the player is riding, the horse is not just a passive mode of transport.

Instead of bucking or biting, a virtual horse glitches, and in that it shows that it is not just an

extension of the player’s body.

Disagreement also give horses back some of their individuality. Agro expresses her feelings in

horsey ways: snorting, neighing, whinnying, tossing her head, refusing to attempt difficult

obstacles, and stamping. She is a friendly presence in emotionally-charged moments, but also

her own being. Agro shows her displeasure by neighing and shaking her head. If she’s really

unhappy (for example, if the player aims their bow and arrow at her), then she will neigh, rear

up, and gallop away. The player does not control the player-character and horse as one, but

rather controls the player-character controlling the horse, and has buttons for reins and

stirrups actions, which Agro can react to.

In Shadow of the Colossus (on the PlayStation 2), the player can call Agro to Wander’s location

using the ⨯ button. Wander mounts Agro using the jump button, △. Once mounted, the left

stick of the controller moves Agro, just as it moves Wander dismounted. The movement

controls for the horse are the same as the movement controls for the player-character. The ⨯

button makes Agro run faster. Dismounted, the ◯ button will cause Wander to pat Agro, if he’s

close enough and he’s put his sword away.

87

In Skyrim, playing on a PC with keyboard and mouse, the player mounts a horse using the left

mouse button, same as they use to act on anything non-violently, whether reading a book,

opening a door, or initiating a conversation. Once on the horse, the player can move the

mounted player-character in exactly the same way as they move the dismounted player-

character – through the W, A, S and D keys. Just as they make the player-character sprint and

jump, Alt and Space make the horse sprint and jump too. Mounted on a horse, the player can

move faster through the world. Once mounted, their player-character also has more stamina

and can keep up a faster speed for longer than they would be able to on foot. However, the

player cannot sneak while they are on horseback, presumably because horses are big, and

they strangely cannot use magic spells while they are on horseback, despite being able to use

two-handed and ranged weapons.

Horse NPCs can’t be discussed without mention of Red Dead Redemption 2. The West of RDR2

is wild and wide, and so the use and maintenance of a horse is strongly encouraged. Horses in

this game have different affective capacities depending on their breed and the effort the player

has taken to bond with them. A racehorse, for example, can run faster and further than other

horses without getting tired, but will get injured and die more easily than a draft horse. A

horse that the player has spent time brushing and patting will refuse to be stolen by NPCs, and

will stay calmer under gunfire. Cool tricks like skidding to a stop or doing fancy dressage

moves are possible if the player has taken good care of their horse or owned it for a long time.

The horse is a character that must be constantly tended to and upgraded, and as such is more

like a virtual pet than the horse-shaped taxis of Skyrim. Players spend a lot of time with their

horses in RDR2, and they can also spend a lot of money buying accessories, medicine and food.

Spending time patting it, talking to it, feeding it healthy snacks, brushing it, and keeping it

clean, or even just riding it around a lot means players can ‘level up’ their bond with their

horse, which widens and changes the horse’s range of actions and reactions. That horses come

in a huge range of types and colours also adds to their appeal as a collector’s item.

Controls in RDR2 are rather contextual, but riding a horse is basically the same as walking a

player-character, but faster. There are some interesting subtleties though. Tapping the increase

speed button (𝖠 on the Xbox controller, ⨯ on the PlayStation controller) in time with the

horse’s footfalls means that the horse will not lose stamina as it runs, and so be able to

88

maintain its speed indefinitely instead of having to slow down to rest. In the settings of the

game, the player can also choose to centre the horse’s controls from the camera’s point of view

(formidable, since there are four camera perspectives), or from the horse’s body. Choosing to

control the horse from the horse’s point of view means the horse moves to its left if the player

toggles the thumbstick left, instead of moving to the left of the screen. The camera settings can

also affect the horse in another way. If the player sets a destination and chooses cinematic

mode, the horse will move on autopilot and the player can sit back and watch as the horse

chooses the best route and pathfinds its way to its target.

The addition of a quarter-ton herbivore to the body of the player-character usually changes

little except speed and the occasional whinny. It is a change to the player-character’s body

rather than a relationship to another being. When Plumwood talks of incorporation, she talks

in a metaphorical sense, where the other can only be understood either as a kind of funny-

looking self (‘they’re just like us really’), or as a foil, possessing all of the good qualities the self

lacks (like in the case of the noble savage) or all of the bad qualities the self defines itself

against (like in the case of the uncivilised colonised) (52). “The human values of humanism are

produced through struggle against the corresponding concept of the animal” as Lindberg puts

it. It has no independent existence, but is defined entirely as the same or the opposite. The

incorporated subject “is construed not as occupying a space on her own account, but as

enclosing a space for another” (52) – it is a boundaryless envelope for ideas about the self. The

incorporation of horses into the player-character, on the other hand, is literal and total.

Whatever the controls for mounting the horse, once appropriately saddled, the player controls

the horse and player-character as one, using the same buttons as they would when controlling

the player-character alone. The horse and the player character become a single being, a kind of

cyborg centaur, and the player has total control. This power over the horse’s body concurs

with the broader pattern of players’ power over others.

RABBITS

[Image removed due to copyright restrictions]

89

Fig. 4: Rabbits frolic in The Long Dark. (author screenshot)

Rabbits are conventional decorations of the open-world role-playing game. Their stilted lope

through faceted forests makes game environments appear more naturalistic. Plumwood’s

conception of backgrounding includes a denial of dependency on the other’s services, but

rabbit NPCs really are just in the background, less a character than part of the woodland

furniture. In their awkward flight they catch the eye of the player, adding interest and

movement to a scene. Since the rabbit is neither friend nor enemy, the player does not have to

do anything about them, and this can be a relief from the logic of other NPC interactions. It can

also be an opportunity for the player to relate to the character in a watchful or playful way, by

following or chasing the rabbit to unknown locations.

The rabbit occupies an interesting place in consciousness. Not fully domesticated, and only

relatively recently brought into service as a pet companion, the rabbit is at once considered

timid and energetic, feral and adorable, communal and inexpressive, dim and successful, easily

born but easily dying. Humans hunt them for sport, farm them for meat and fur, sacrifice them

for religious services, keep them as pets, persecute them as pests, experiment upon them in

clinical trials, admire them in the countryside, and endlessly pontificate upon them in

children’s books (DeMello 10). It is no surprise then that they should show up in video games.

In fact, the rabbit is a common figure in a large number of games, where it can be chased, shot

and eaten (Skyrim), skinned and used for crafts (Horizon Zero Dawn and The Long Dark),

collected (Stardew Valley and The Legend of Zelda: Spirit Tracks), helped (Animal Crossing:

New Horizons) and thought of as cute (all). The paradoxes of the digital rabbit reflect the

twists and turns of our shared history, in which the rabbit appears both as a wild animal with

many useful purposes, and a creature of imagination and repeated representation.

As in the case of chickens, the perceived benign timidity of rabbits is an opportunity for some

switcheroo humour. As a non-predator, the rabbit fills a character role as someone who is

nervous, needy, fearful, but also friendly and enthusiastic. A rabbit who looks cutesy but acts

ferociously is so common so as to be predictable. It seems fair to say that the more adorable a

bunny seems, the more likely it is to rip your arms off. Like Monty Python’s killer bunny of

90

Caerbannog, aggression from rabbits is seen as incongruous and therefore funny. Rabids, the

antagonists of the Rayman series and stars of their own spinoff series, are violent but

cartoonish creatures whose poorly thought out aggression results in much slapstick humour.

In Braid, rabbits are a formidable enemy, though their innocuous appearance is even enhanced

by their disguise as meowing flowers. Cave Story and Terraria feature bunnies that are

stereotypically cute until environmental factors (a certain food, a certain moon phase) turns

them into ravenous hellbeasts.

Echoing Berger, Plumwood writes “The master’s view is set up as universal, and it is part of

the mechanism of backgrounding that it never occurs to him that there might be other

perspectives from which he is the background” (48 emphasis in original). As part of the game

environment, the rabbit is an inconsequential character whose existence or nonexistence has

only a tiny effect on the player’s story. But as part of the background, they create the space for

a player to exist; without a game environment there is no player. The player is just as

dependent on the game environment as the environment is on the player – and even moreso.

“The master more than the slave requires the other in order to define his boundaries and

identity, since these are defined against the inferiorised other” (48 emphasis in original)

Plumwood argues, and this implies that the power of the player is a result of the game’s

environment, including rabbits, being pretty and forgettable. For the player to be important,

their surroundings must be unimportant. Rabbits, as a synecdoche of nature, have to be

backgrounded through “mechanisms of focus and attention … so the denied areas are simply

not ‘worth’ noticing” (48). The decision to make rabbits decorative snacks is a way of making

the player-character the centre of the game world, by making everything else peripheral.

CATS

[Image removed due to copyright restrictions]

Fig. 5: The slugcat of Rain World in its natural habitat (lizards visible bottom left and right).

(author screenshot)

91

In this section we’ll be cheating a bit. The best games featuring cats present them as player-

characters, and their position as small, independent, yet vulnerable mammals shapes their

relationship to the world and its inhabitants in an interesting way. In both Rain World and

Alley Cat, players play as small cats who have to navigate big worlds. They are predators to

some and prey to others. On the face of it, this doesn’t sound so dissimilar to player-characters

who attack and are attacked by enemies. Where other player-characters are central and

powerful, though, the cats of Rain World and Alley Cat are disposable pets who have little

control over their worlds. Their affective capacity is small, and their most useful ability is to

run away fast. A small demonstration of this decentring is the cameras of both games, which

are fixed on rooms rather than focused on the player-character. The cats of both games move

about these rooms but are not followed by the camera, and each room is a tableaux of

collected elements, rather than a portrait of an individual.

Rain World’s protagonist is an omnivore, and is positioned in the middle of a food chain. This

fact shapes the player’s attitude and relationships towards other animal-shaped beings in the

game. If the player were a carnivore, this would result in more hunting than foraging. If the

player were a herbivore, this would result in more exploration than killing. Being omnivorous

means the player can take a mixed approach, or can choose one or the other playing style.

Many games take a linear view of relations – some creatures are enemies, enemies attack on

sight with violence, and that is just the way of things. It is the story of a powerful being

constantly defending its precarious position, forever teetering on the pinnacle of success. But

an omnivorous approach, such as that taken by Rain World, positions the player at one node of

an interdependent network, an ecosystem that cannot be conquered but can only preserved in

constantly fluctuating equilibrium. Rain World’s protagonist, a medium-sized mammal, is prey

to the lizards, the birds, the bugs and the carnivorous plants, but predator to the bat, the flying

squid, and the moth, as well as various fruits and fungi. The NPCs of Rain World also have

relations of their own: centipedes will prey on lizards, squid will eat bats, and carnivorous

plants and vultures will attack anything that moves. The aim of the player is to figure out how

to survive in this complex web of relations, while also navigating a dilapidated maze-like

environment. Unlike many post-apocalyptic exploration games, the slow uncovering of a tragic

history is not a central part of the gameplay, which instead focuses on coming to terms with

this present system of relations.

92

Rain World’s NPCs would be fine if the player-character didn’t exist, and could happily live out

their existences in a screensaver. They do not yearn for meaning or for anyone to arrive and

make their world better. They can affect each other just as well as the player-character can.

Every being in the game strives to preserve itself while constantly changing as a result of

participating in a network of relations. In this, everything in Rain World can be said to

constantly be attempting to individuate, rather than be an individual with finality. Rain World’s

network of relations reveals the affective nature of networks, in which beings try to continue

being themselves even though they are changing all the time. The lizard is always lizard-

shaped, but this does not mean it can’t be a friend, an enemy, or food to the player. There are

no essences outside the thing and its relations, because what is good is always changing. In

Rain World, what is good or bad is always what is desirable and undesirable in the moment.

[Image removed due to copyright restrictions]

Fig. 6: The cat of Alley Cat warily watches an enemy broom from the safety of a table. (author

screenshot)

1983’s Alley Cat is a simple game in which players must catch mice, fish, birds, and feelings,

while avoiding spiders, brooms, electric eels, dogs, and other cats. The levels of the game are

arranged as rooms of an apartment block, and choosing a window to enter is a challenge in

itself, as tasty mice scurry along clothes lines above, enemy cats pop out from rubbish bins

below, humans throw boots in between, and bulldogs run wild underneath. Life is quite hard

for the small black cat, especially if the player doesn’t have cat-like reflexes themselves.

NPCs are threats or achievements for the cat of Alley Cat. As a feral cat simulator it is frenetic,

offering a world where co-existence is uneasy, but possible if the player moves fast enough.

The player finds little solace amongst those of their kind; they are all competing for the same

prizes. The NPCs of Alley Cat do not encounter the same environment as the player-character,

however. In the romance level, the player-character’s rivals do not seek to reach the love-

interest first, but endlessly patrol back and forth on ledges beneath it. The clothesline mice do

not get hit by flying boots, and dogs are not swept up by brooms. The player-character cat is

93

uniquely besieged. The animal characters of Alley Cat are not irrational, but nor are they

human. Instead they are all defined by their precarity, their vulnerability to their world.

Alley Cat does not decentre the player-character exactly, but it does decentre the human. Not

by having a cat as a player-character, but by having the player-character as a creature swept

up in, rather than a direct cause, of affect. The cat of Alley Cat both deals violence against mice,

and has violence dealt against it. As the game progresses though, the simulation becomes

unhinged, as the mice run faster, the brooms sweep ferociously, and more and more dogs rush

through the apartment building. The cat does not gain any powers, but just has more unique

opportunities to be overcome. Progress for the player is measured in possible ways to die. The

player must be more and more careful while the world becomes ever more unpredictable and

wild. The player has less and less humanity – “the power to shape and break the lives of

others” (Lindberg).

Rain World’s sense of verisimilitude comes from both the player-character limited capacity to

affect, and the nonplayer characters’ comparatively substantial capacities to affect and be

affected. Alley Cat, while not quite as complex, also demonstrates how in-betweenness and

relative powerlessness can function as a mechanism for decentering the human.

WILD ANIMALS

[Image removed due to copyright restrictions]

Fig. 7: The creature of Ball, and its ball. (author screenshot)

The creature in Vectorpark’s Ball is an extremely simple being that loves playing with its ball.

The creature has no real-world analogue: it looks a bit like a floppy baseball bat with a

propeller where the handle should be. It moves by propelling itself along the ground, barrel

first, looking a little like a dog snuffling along the ground. It can rear up on its propeller to

throw or catch its ball. The creature, which we’ll call Bruce for simplicity’s sake, and its ball,

94

are the only things in their world, but Bruce doesn’t seem to mind. It is totally absorbed with

its ball, rolling it around, balancing it on its tip, throwing it around and racing to catch it again.

It does not acknowledge the player. The player cannot acknowledge it either – clicking on

Bruce does not produce any noticeable effect. In this tiny, first-person interaction, the only

thing the player can affect is the ball. Both Bruce and the player are united in affecting this love

object. The player can pick up the ball, move it around, and bounce it high, but the effect is not

like playing with a dog. Bruce waits patiently while the player has the ball. It does not seem

anxious or excited. It just waits for the ball to come back. If the player releases the ball, Bruce

starts again where it left off. Throwing the ball elicits a similar reaction: Bruce trundles over to

where the ball ended up, and starts playing by himself again. Attempts to excite Bruce by

holding the ball over its head only induce Bruce to fruitlessly circle under it, reaching

upwards. Bruce never invites the player to interact, but also never seems resentful that the

ball has been tampered with. The player can never benefit or harm Bruce, just interrupt it. If

Bruce has needs, it won’t ask the player to fulfil them. If Bruce experiences pleasure of pain, it

won’t share them. Unlike real or represented dogs, Bruce does not need a human to act on its

behalf; it will never be a companion.

[Image removed due to copyright restrictions]

Fig. 8: The player moves the ball of Ball, and the creature moves to its position. (author

screenshot)

The mechanics of Ball are extremely simple: the player can move the ball around. The world

has gravity so a dropped ball will fall, and the world has a ball-lover, so the ball will be loved.

Spinoza says of ball-lovers “love is nothing else than the idea of pleasure accompanied by an

external cause” (140), but he acknowledges that “if anyone conceives that an object of his love

joins itself to another with closer bonds of friendship than he himself has attained to, he will

be affected with hatred towards the loved object and with envy towards his rival” (153). One

can feel deflated when they cannot effectively act, as “when the mind regards itself and its own

power of activity it feels pleasure; and that pleasure is greater in proportion to the

distinctness wherewith it conceives itself it its own power of activity” (165). Technically Bruce

is not so different from an NPC that has been programmed to follow a mouse cursor, but the

95

fact that it does not follow the player’s mouse and instead responds only to the ball is

confusing. It does not act how an NPC is expected to act, and furthermore, nothing the player

do can make him express pleasure or pain. In fact, the player mas as well not exist for Bruce.

Bruce does not exist for the player, but nether does the player exist for Bruce. Bruce does not

exist for the player because it is an NPC, a fictional being that exists only for a short time as

pixels on a screen. But Bruce does not exist in another sense: it cannot be affected by the

player, the only thing the player can interact with is the ball. Bruce and the player are on either

side of a vast chasm of incomprehension. The player is not the centre of Ball, the ball is. The

player is decentred in favour of a toy, and that is a disorienting experience. There is nothing

the player can do in Ball, and more importantly there is nothing they should do. They cannot

do anything good, and they cannot do anything bad. They can only open and close the game,

affording Bruce and the ball the experience of existing and being affected by each other.

Bruce is a wild NPC: it does not subsist on any human attention except the most passive. It is

independent in the context of gameplay. But as Plumwood asserts, conceptions of wilderness

should not rely on defining autonomy as hyperseparation (163), and wildness does not mean

complete isolation from humans. Bruce still only exists when a human boots up the game.

Plumwood suggests that wilderness should be seen not as a place of exclusion but a domain

where everything is autonomous and sovereign, “free to work things out according to self-

determined pattern, which may be those of sameness or difference” (163). It should be a place

where “the self does not impose itself … visitors … must see themselves through the other’s

eyes, must bend themselves, as is appropriate for visitors, to the other’s ways” (164). In other

words, a wild game would involve letting all the NPCs just get on with their own lives.

CONCLUSION

Humans in games are generally disappointing. They are shaped like people but their

limitations quickly become obvious. The popular image of animals as simultaneously aloof,

slaves to programmed instinct, unknowable, machine-like, wild and free, and resolutely other,

96

means that animals modulate well to the digital realm. Animals are already expected to be less

predictable, intelligent, active, and creative than humans, so the same qualities that leave

human NPCs unsatisfying do not have such effects on animal NPCs. This characterisation has

two effects: compared to human NPCs, animal NPCs are both easier to use as a resource, and

harder for some to kill.

Relationships between humans and animals are economically, ecologically, ontologically and

individually significant. Even when individual humans and animals are not in direct contact,

human-animal relationships are systematically at work behind the scenes. It is impossible for

a human to move in the world without encountering animal bodies in some form or effect, and

vice versa. In games, animals do not have the same sort of bodies, but designers still seek to

replicate their affective power. Animal NPCs epitomise something about all NPCs. Their bodies

are the key to their role, and they exist because they are useful. They are made to do work, but

only certain kinds of work. They provide ecosystem services, but in an economically classic

way (Chang 56). They are production-line robots, but instead of manufacturing cars or

biscuits, they are made to manufacture affect.

Berger states that humans increasingly experience animals through media representation, and

that this has certain effects on our subsequent attitudes towards non-human others (Berger

2009). In short, Berger argues that the way we view animals affects the way we view animals.

More accurately, Berger argues that humans increasingly experience representations of

animals instead of actual animals (DeMello agrees 58), and that representations of animals are

no replacement for the role animals used to perform in the formation of the human self.

Captured in the gaze of an incomprehensible being, the human is an environmental object to

the animal’s subject, “he is being seen as his surroundings are seen by him” (Berger 5). When

a human is seen by an animal the human is subjected to the same gaze as the human

themselves subject others to. But Berger forcefully argues that this way of being has been

erased in the 20th century – animals have become (seen and used as) machines, then

resources, then images. Images of animals cannot look back at the viewer, and so the human in

this encounter has no experience of being in the presence of something non-human. As

represented animals proliferate, the human only becomes surrounded with more human

things, as Briggle notes. The represented animal ceases to be a real other, and is folded into

97

conventional narratives of the family (as a pet), science (as a case), commerce (as a

commodity), or the exotic (as a curio and necessary counterpart to normality). The animal

becomes a domestic subject, a collection of data, a resource of meat and pelt, or a repository of

hopes about a sublime wild, rather than a being in itself. Games, then, “exaggerate and

formalize problems of mastery, problems negotiated in each instance by the player”

(Lindberg).

Animals can’t talk (DeMello 19), and for that they “exist as mirrors for human thought; they

allow us to think about, talk about, and classify ourselves and others” (16). Representation of

animals draws heavily on other representations of animals, which Baker notes has some

implications for real animals (197). As such, ideas about animals, what they symbolise, how

they’re used, what character traits they stand for, are even more important for the design of

animal NPCs as their real counterparts. Animal NPCs are wrapped up in mythology and

cocooned in presentiment, only to emerge as very normal game characters. As bearers of

morality (Burt 208; van Ooijen 40), animal NPCs elicit similar calls for concern about

computer game violence as human NPCs, but it is interesting to note that many players will

avoid killing animal NPCs when they don’t avoid killing human NPCs. While lovely, this

suggests that animal NPCs are thought closer to animals than human NPCs are to humans,

which isn’t necessarily great news for animals. Animals are closer to animal NPCs, but that is

because both are made comparable, by being other to humans.

Animals are pretty cool, and they’re pretty cool in games too. While they can play the role of

an enemy or friend, they offer a third basic way to relate to NPCs: as a consumable resource,

and as an other. While enemies are usually positioned as faceless mooks and friends as

inferiorised minions, animals inhabit a fuller version of the concept, acting as background to

the player-character’s foreground, masses to the player’s individuality, strangeness to be

explored, tools to be used, and flesh to be eaten. They are the epitome of NPCs because they

explicitly exist to be used.

Animal NPCs provide the player with things they can’t get from human NPCs. They also give

the players extremeness – extreme loyalty, usefulness, and prettiness. At a minimum, pets

need food, water, shelter, affection, and some training. Computer game companion animals

don’t even need those. Virtual farm animals easily and painlessly gift the player meat, milk and

98

eggs. Horses combine with the player-character to create a being greater than the sum of its

parts. Rabbits are very nice to look at and help to create a lovely pastoral atmosphere. Robot

dinosaurs aren’t really alive, so it’s fine to kill them. PinX atas are cute and voluntarily decide to

live in the cushy confines of the player’s well-run zoo. In one sense, this is an idyllic world. It

also reflects pretty pernicious ideas about real animals, their place in the world, and of

humans’ place in the world. These unoriginal and simplistic ideas about how animals work

contribute to uninteresting game design.

New narratives about selfhood and otherness, and new positions in relation to NPCs, can

make better games and better NPCs. “The player-centric nature of video game design

simulates anthropocentric thinking and opens space to interrogate a particular imagining of

humanism through the figure of the player, an interrogation principally structured by the

same human/non-human (animal) logic that underlies humanist dualism”, as Lindberg puts it.

Rain World’s interdependent ecology, Alley Cat’s surreal simulation and Ball’s aloof companion

demonstrate easy, low budget ways to put players in weird decentred positions. In the final

chapter of this thesis, we will look at some more games that attempt just that.

99

CHAPTER THREE: ENEMIES

INTRODUCTION

Unlike in the real world, it is easy to identify and defeat enemies in a computer game. This is,

in fact, part of their appeal. Enemies are a staple of the adventure and role-playing genres of

games precisely for their abilities to provide simple, thrilling danger and reinforce the player’s

place in the world by being a complementary counterpart. Game enemies are generally very

effective characters. They are hateful and frustrating, and seemingly feel the same for us.

Enemies cannot be dissuaded from their obsession, nor can they be reasoned with. Often, only

death can quell their furious antipathy towards a player. Their hostility is essential, their

badness innate. Enemies are creatures it’s always OK to hurt.

The player is always a threat to an enemy; the enemy always sees the player as a danger. To a

player, an enemy can be a chore, a fun way to pass the time, a duty, a harassment, a challenge,

or a tormentor, but to an enemy character, the player is always an emergency. Players affect

enemies by touching their body to the point of overwhelm. Through feelings of achievement

and overt rewards, players affect themselves as much as they affect enemies; what enemies

lose players gain. When the player fights an enemy the player has a good chance of winning,

though it might take a few rounds of dying to learn the techniques. Players, then, are rewarded

for affective work. Enemies affect players by reminding them of things, threatening them with

loss (of time, place, accumulated wealth, or esteem) and touching the player-character. They

are made to be overcome, to be able to be affected to death, and despite their belligerence they

are willing participants in this process.

The affective capacities of enemies (their ability to affect others and to be affected themselves)

are great, and range from inducing pre-individual wariness in the player to dropping gold

coins dropped after they die. Players seek to gain something out of interactions with

nonplayer characters, and enemies in games are always enemies-with-benefits. When the

player defeats an enemy they receive a reward, primarily the reward of not having to fight any

100

more (though sometimes, like in Shadow of the Colossus, rewards are complicated). An

enemy’s difficulty is generally reflected in the quality or quantity of rewards given to the

player upon winning. A player can expect a hefty compensation for a major battle, which can

be incentive to start a major battle. Material acquisition and progress in the game, then,

depends on overpowering many enemies before the game is concluded. Some games include a

tally of how many enemies have been killed, for the purposes of affecting the player with pride

or remorse.

Games would be much easier without enemies at all. Sonic would be able to run infinite loop-

de-loops, the slug-cat of Rain World could collect subterranean fruits at its leisure, the race-car

of Speed Race would never be run off the road. Defeating an enemy results in the removal of an

obstacle and the continuation of the player’s journey. That is, enemies are not part of the

player’s story but block the story. They are outsiders interrupting the flow of an ideal game,

even though interaction with them may take up most of the time of playing. Winning against

an enemy returns the world to order. Most games would also be a little dull without enemies.

Aloy would wander around a perfect Eden, Chun-Li would impatiently stand around, and

Metal Gear Solid would take ten minutes to complete. As well as loot and a challenge, game

enemies add meaning to a player’s experience of a game.

An enemy is relatively easy to define. Enemies are nonplayer characters controlled by the

computer who oppose the player’s character. While earlier typologies of nonplayer characters

differentiated between nonplayer characters and ‘monsters’ (very simple enemies), this

distinction is no longer useful (Warpefelt ‘thesis’ 38). Mostly physically and violently, but

sometimes, like in the case of the Portal series, through words and structures. Enemies seek to

cause the player harm, and enemies exist to be defeated by the player. They are easily

identifiable, having been designed with unique appearances and distinctive silhouettes for

quick recognition. As well as opposing the player character in battle, they often also oppose

them in style, being dark where the player is light (Rain World), nonhuman where the player is

human (Horizon Zero Dawn), or large when the player is small (Shadow of the Colossus), for

example.

101

Warpefelt notes that enemies “will generally attack the player on sight, and in some cases they

will even seek out the player to attack them (common in RTS [real-time strategy] games). The

player should be able to dispatch them fairly easily, although not too easily … In most cases,

NPCs who are primarily of this role will be portrayed as actual faceless enemies. … In cases

when they are not actually masked, their appearance will be very generic, often to the point of

being clone-like. Furthermore, they will be equipped with indicators as to their role or

profession, uniformed soldiers with weapons or monsters with fangs and claws.” (Warpefelt

‘thesis’ 88). Isbister argues similarly, saying, “Enemies seek to destroy the player and vice

versa. Enemies usually do not have the strategic power of the player, although they may have

equal or better fire power. Enemies are the antiplayer equivalent of minions—they do not

require much time and emotional investment from the player, and their contribution to game

play is usually to be cannon fodder. Enemies are usually dehumanized (portrayed as aliens,

faceless nameless soldiers, or crafted in a cartoonlike way), allowing the player to treat them

as an “other” and not to become upset by their deaths” (242). They act as barriers to player

progress, obstacles to be overcome, gatekeepers to new and better circumstances, challenges

to be met. The player is always the centre of these desires. It is only in ending that they finally

find their purpose; it is only when meeting the player their life finds meaning. They are simple,

and the player’s duty towards them is simple.

When the player dies, they are usually reborn, but when an enemy dies they are dead forever.

Luckily, enemies are also rather generic, and look-alikes abound. To die, for an enemy, is to

disappear, or to disappear and be replaced with a less functional representation of their own

dead body. To be respawned is to be reborn on the player’s terms, either when the player-

character dies or the player goes back to a save-point. The life and death of a game enemy is

Sisyphean, if Sisyphus’ rock was fighting back.

Special, strong enemies called bosses live at the end of levels and in important locations, and

can be difficult to defeat. These characters are individuated from other enemies in skills and

appearance. Encountering a boss is eventful, an indication to the player that they have reached

some sort of climax. The boss, however, is not so completely unique that the player doesn’t

know what to do with it. Through repeated interactions with simpler enemies, the player has

learnt the techniques that will help them overcome this new character. The novelty of the boss

102

teaches the player more about the game world (especially the power structures of the game

world), and the player’s position within it. Beating a boss solidifies a player’s growing sense of

mastery, while losing against a boss highlight’s the player’s place in the game’s hierarchy.

Game enemies are not really similar to anything humans encounter outside of games. Their

instant, furious hatred, their willingness to fight to the death on sight and their unwillingness

to negotiate is not something encountered in everyday interaction. Yet relations with game

enemies are also paradoxically easy to intuit and participate in. The player also becomes a

game enemy, attacking without provocation, fighting to the death and preemptively rejecting

potential deescalation tactics. This kind of unique interaction with nonplayer characters, this

unique relationship between player and enemy, continues to be a common, enjoyable and

celebrated part of computer games.

In this chapter, this thesis will use 'enemy' as a shorthand for 'enemy nonplayer character'.

THE ENEMY, A NOVEL AND SUCCESSFUL RELATIONSHIP UNIQUE TO

COMPUTER GAMES

Enmity reportedly existed before the invention of the computer game. It is, according to some

commentators, a noble and natural pursuit, but also something that can easily corrode a moral

character (Spinoza 218). However, the form of enmity encountered in computer games by

computer game enemies is a novel and unique relationship which has no equivalent in the real

world. The enmity of the game enemy is an idealised antagonism, the fantasy of a military

enemy or of animals in a fighting ring. Enemies are representative of dogs, foreigners,

necromancers, orcs, and lovecraftian monsters, but their real referent is the idea of pure

opposition itself.

Usually, people walking down the street aren’t attacked by dogs, beset by mafia goons,

targeted by zombies or bitten by little walking mushroom people, and they do not have to beat

103

these things to death in order to go about their day. But these occurrences are ubiquitous in

computer games, and, strangely, they are intuitive to manage. A computer game enemy is both

something completely outlandish and something the player knows the right way to react to.

Enemies are not like anything a human encounters in everyday life, and the way a player

reacts to an enemy isn’t based on any reality either. Humans, like other animals, prefer to

avoid and deescalate threatening situations rather than rush headlong into them. But in a

computer game, players and nonplayer characters look for trouble and relish a fight. This

makes the human-enemy relationship one of the most unique, paradoxical, and intensely

affecting relationships found in computer games.

1971’s Computer Space, the first computer game to include an enemy nonplayer character, was

essentially a one-player version of the 1962 two-person spaceship combat game Spacewar!. In

Spacewar!, two players control two spaceships, and must dogfight to death in the gravity well

of a star. In Computer Space, the player’s rocketship fights two UFOs, both of which are

controlled by the computer. The movements of the UFOs are not particularly sophisticated,

making these nonplayer characters more akin to fairground wooden ducks than the enemies

we know and loathe today. They were a technical challenge, a way for the player to prove their

skill rather than an other to be empathised with. But the idea of the computer standing in for a

competitive human was set. The invention of the enemy nonplayer character, then, was as a

replacement for a human friend.

Early computer games were most often multiplayer games based on sports or battles. The

enmity present in them was a familiar enmity, present on the sports field or across a tabletop

boardgame. While relationships between players in these arenas can still be fraught, and the

playing of a game can be a subtle opportunity to establish or reinforce a broader pattern of

dominance (as merrit k points out), enmity between players in other contexts is more often

temporary, purposeful, competitive, and forgettable. The enemy of an early computer game

then, is not in a class by itself, but rather akin to the existing human relationship of play

competitor.

But game enemies soon diverged from being a replacement for a human companion. The

unpleasant and unbeatable game Marienbad demonstrates how computer enemies can easily

104

become ruthless. Based on nim, an ancient logic puzzle, the two players of Marienbad take

turns picking up virtual sticks from a digital pile. The player left with the last stick is the loser.

The game Marienbad is designed to be unbeatable, with the computer opponent always

making perfect calculations and moves, and a single mistake by the human player resulting in

their losing. While not a character as such, the computer is a competitor, one as single-minded,

exacting, and merciless as enemy nonplayer characters.

These qualities then, being ruthless and being an object against which players can

demonstrate their skill, have continued to be key properties for enemy characters. Enemies

today still need to be as cruel and as challenging, but do not need to be substitutes for human

acquaintances. The enemy relationship has evolved into a new thing in its own right. It has

intensified some ideal aspects of human competition (rivalry, difference, antagonism,

pageantry), but has dropped others (camaraderie, sportsmanship, temporariness, extra-game

commonality). Enemies, then, are unique because they are a certain type of relationship

(sporty or competitive), applied to a different context (a medieval village; a mafia-ridden

town; a cyberpunk dystopia, etc).

Game enemies are entirely a product of the single-player computer game. In early multiplayer

games, humans could be each other’s enemy, but in a single-player game, the computer has to

pretend to be a friend pretending to be an enemy. As other people are always reminding us,

computers are imperfect substitutes for friends, and in attempting to fill an imaginary human

absence, the computer instead becomes a real inhuman presence. Without another human to

fight against, the computer steps in, but computers fight differently. Through iteration, the

solidifying of convention, the pursuit of novelty, and programming limitations, these

differences compound over time.

However, the enemy relationship shows that nonplayer characters can be affective without

having to closely represent anything real. Since enemy characters look and act differently to

anything a player has encountered before, it seems like they should feel strange and

uncomfortable. But there are a number of reasons enemies seem strangely familiar. Game

enemies are easily understood because power is an intrinsic part of everyday life. Though

random acts of violence might not be too common, power dynamics and the perpetuation of

105

hierarchies is pervasive. Games are also a highly intertextual medium, and reference each

other frequently. Familiarity with one game carries over to other games, to varying extents. A

brawler like River City Girls makes more sense if the player has played any earlier games of the

Kunio-kun series, any brawler games, any fighting games, or any games at all. The simple

enemy character can be found in many games, so even though the player might be unfamiliar

with the painted troll of Oblivion, they still get the gist if they’ve played other games before.

Each game also has some sort of learning curve, even if it’s s steep one. Games introduce

players to enemies, which serves as a tutorial and also as a measure of progress. If a player

stops getting attacked by rats and starts getting attacked by goblins, they can be sure that they

have levelled up. Player progress measured against a hierarchy of speciated enemies points to

another basis for easy enemy relationships.

ENEMIES OF PHILOSOPHY

Game enemies are philosophically interesting as they are both creatures for which there is no

real world equivalent, and creatures that invite a familiar relation of domination. Although

they act towards the player like no other being, the ways they are designed and the ways in

which the player views them is well-established. This is not just a case of enemies being

representations of stereotypes of human minority groups or savage wild animals, though that

certainly plays a part. The key to enemies’ familiarity is the position the player takes in

relation to them. A position that requires enemies to be underneath the player in skill and

importance, to be repelled by the player as an inferior and absorbed into the player as a tool.

The process of this domination, of this struggle for power, is already codified in other contexts.

Even if the player wanted to have a different relationship with an enemy, this is usually not

possible.

For Spinoza, good and bad are never objective qualities of a thing but always good for some

purpose and bad for some purpose, as “the terms good and bad … indicate no positive quality

in things regarded in themselves, but are merely modes of thinking, or notions which we form

from the comparison of things with one another. Thus one and the same things can be at the

106

same time good, bad, and indifferent” (190). So if the player wants to continue progressing in

a game, game enemies are bad because they could prevent that desired outcome, but if the

player is tired and looking for an excuse to stop, or is seeking excitement, or wants to show off

their fighting skills, game enemies might be good as they help the player achieve those aims.

Taking the first case, in a Spinozan view of enmity, enemies are bad because they counter the

player's desires, particularly their desire to persist in the game world. However, the

defeatability of enemies repeatedly challenges and fulfils this desire, with the result being that

the uncertainty of possibly losing everything is essential to the joy of normality being

reaffirmed. The enjoyment of a game containing enemies is that it could stop at any moment

(but doesn’t).

However, enemies can be problematic for Spinozans for the emotions they elicit in players. An

enemy can be the source and object of feelings of anger, hatred, revenge, envy, and derision –

emotions which, Spinoza notes, all arise from the experience of pain. Pain is always to be

avoided, as it is a sign the body has been made less capable of action (138). But hatred for

game enemies is somewhat inevitable, as “He, who conceives himself to be hated by another,

and believes he has given him no cause for hatred, will hate that other in return” (156). A

Spinozan does not want enemies, as the relationship and the emotions that arise from it are

painful and so inherently detrimental to one’s wellbeing. Even triumphing over an enemy, a

happy occasion, is tinged with sadness, as “joy arising from the fact, that anything we hate is

destroyed, or suffers another injury, is never unaccompanied by a certain pain in us” (160)

(although it does feel pretty good (144)). Furthermore, emotions arising from desire rather

than rational examination aren’t that great either, as “[hope and fear] show defective

knowledge and an absence of power in the mind; for the same reason confidence, despair, joy

and disappointment are signs of a want of mental power. For although confidence and joy are

pleasurable emotions, they nevertheless imply a preceding pain, namely hope and fear” (220).

As to whether killing enemies in computer games is bad, Spinoza suggests that “a given action

is called bad, in so far as it arises from one being affected by hatred or any evil emotion. But no

action, considered in itself alone, is either good or bad, one and the same action being

sometimes good, sometimes bad” (227). So Spinoza can be said to not be a fan of enemies, not

because killing them is problematic in any way, or the player behaves badly, but because any

107

experience of pain is harmful, even in the experience of play, and being a rational philosopher

is the most fun you can have anyway (270).

Enemies could also be problematic for Spinozans as good people should want the best for

those around them (211), and should have desires that serve their community as well as

themselves. Spinoza makes it clear, however, that computer game nonplayer characters are not

part of his moral community, as “besides men, we know of no particular case in nature in

whose mind we may rejoice, and whom we can associate with ourselves in friendship or any

sort of fellowship; therefore, whatsoever there be in nature besides man, a regard for our

advantage does not call on us to preserve, but to preserve or destroy according to its various

capabilities, and to adapt to our use as best we may” (241). When Spinoza specifically states

that in his propositions “Hatred can never be good” (218) and its corollary “Envy, derision,

contempt, anger, revenge and other emotions attributable to hatred, or arising therefrom, are

bad” (219) he is only talking about hatred towards humans (219). In his exhortation that to

“make use of what comes in our way, and to enjoy it as much as possible … is the part of a wise

man. I say it is the part of a wise man to refresh and recreate himself with moderate and

pleasant food and drink, and also with perfumes, with the soft beauty of growing plants, with

dress, with music, with many sports, with theatres, and the like” (219), it seems computer

games would fit neatly between sports and theatre. However, love can be excessive (218), so

rational people should not become so enthusiastic about one love object that they neglect

other parts of their life or find themselves limited in action because of it.

So whether game enemies are definitively good or bad cannot be answered in a Spinozan way

– once again it is on a case by case basis. Insofar as computer games are fun entertainment,

they are a good thing. If a player finds themselves loving the game excessively, or feeling pain

during the game, it can be harmful. If a game enemy’s destruction of the player character is

entertaining or joyful, it can be an overall good thing for the player. If the destruction of the

enemy is poignant or saddening, it can be a bad thing. And the same interaction with the same

enemy can affect different people in different ways, depending on their history and what other

images they are reminded of when they associate with the beings. When it comes to affect,

Spinoza reminds us that “different men may be differently affected by the same object, and the

same man may be differently affected at different times by the same object” (163). Spinoza’s

108

insistence on particularity and contextuality is his strength. Things aren’t just good or bad,

they can’t be generalised or categorised, because individual elements can’t be plucked out and

judged singularly from the whole totality of swirly cyclical affect.

Plumwood, on the other hand, is up for a fight. Her work on domination gives us an

ecofeminist insight into the workings of the player-enemy relationship, and the familiar

foundations of a strange situation. Plumwood’s perspective helps to unpick the non-game

origins of human-enemy relationship dynamics, and how the player and the enemy’s pre-

individual view of each other shape how they will act. Though the logic of domination is

present in all interactions between players and nonplayer characters, in enemy relationships

we find the logic most obviously displayed. It must be stressed that neither nonplayer

characters nor players are particularly harmed by this dynamic (except for the lost

opportunity of trying something new in game design).

A game is a place where some beings are better than other beings. Better at fighting, better at

jumping, better at collecting, better at running, better at being a real human being. A game is

also a place where the player has more or less power than other characters; they are always in

relations defined by power. But while those other characters are static, the player progresses,

gaining power until they are at the top of the game's hierarchy. This betterment in status is

measured on leaderboards and speedrun times, but it is also visible in the types and

frequency of enemies. As the player becomes better at fighting, more challenging enemies

appear, and enemies that were once threatening become trivial. In this way, the player can

reflect on their place in the world and observe their relative status in the types of others they

have to deal with. Where once the player struggled with overcoming mudcrabs, now they

overpower demons.

Hierarchies of being, Plumwood observes, tend to be constructed and justified in similar ways,

no matter what the beings are. When beings are made superior and inferior to each other, the

same dynamics are recycled. The superiority of players in games seems natural, but is

constructed. As enemies are designed to oppose the player, often in multiple ways, they

demonstrate this process clearly. In the logic of domination, there is only one way to be a good

person, and all other ways of being fall short. Those who fail to meet certain standards either

109

aren’t good people or aren’t people at all,. This state of affairs appears to be inevitable and

natural, but is constantly policed (32). Humans, men, colonisers, or in this case players, are

held as the standard that others must aspire to. They are said to possess certain qualities –

rationality, civilisation, agency – and these qualities justify the hierarchy. To be fully

considered as a moral agent, worthy of being cared about, one must be as close as possible to

this ideal person. These are fine qualities in themselves, Plumwood notes, but are not really

exclusive to any kind of person and are also not so special that they justify domination (24).

Dominated groups are “thrown into” alliance not because they have some essential similarities

but because the mechanisms of oppression work similarly (21). For things to be recognised as

worthwhile people in Western societies, Plumwood argues, they must fit a rigid definition that

has been developed by centuries of philosophers but also suspiciously resembles those

philosophers.

Nonplayer characters in computer games obviously aren't good humans. They are not rational,

independent, or masters of their world. They’re not even solid or conscious. But Plumwood

argues that beings shouldn't have to be like the best kind of people in order to be cared about.

Game enemies are designed with dualisms in mind. They are large to the player’s smallness,

inhuman to the player’s human, dark to the player’s light, and otherwise characterised by

opposition and dissimilarity. Dualism, Plumwood notes, erases continuity and asserts

essential difference. There is no commonality with the enemy, and so no hope of

reconciliation. Dualism constructs the game enemy as an other, capitalising on existing

difference (humans as autonomous and rational, nonplayer characters as automatons and

mindless) and creating difference (in character design that makes the enemy look and behave

very differently to the player). The otherness of nonplayer characters is real, but it is also

further exaggerated for effect. In terms of enemies, this means that their machine-like selfless

relentlessness is emphasised, and it is the player’s humanness (in their ability to learn and

play artfully) that results in the player’s success. For enemies to become something more

complicated than not-players, the answer lies not in trying to make them more human, but in

showing the ways in which both players and nonplayer characters are important beings in the

context of the game, do not have such strictly defined roles, are interdependent, and are

different from each other as well as the player (see Plumwood’s list of the structural features

of dualism 60).

110

Enemies could easily be made unbeatable, if anyone wanted such a thing. The defeatability of

enemies, though, is the point. They have to be able to be mastered. Even if the player never

achieves it, they know it must be possible and it’s only their own performance which is

lacking. Beatability is a trite example of how enemies are made inferior. Even if they struggle,

players know that enemies have been designed to be overcome.

As man-made objects, Plumwood would argue that nonplayer characters don’t have it too bad

and don’t really need to be liberated (as per her discussion of Coke bottles 138), but her

analysis of power relations is still useful both for an analysis of how power works in games,

and a framework for how different relationships can be designed. Her explanation of how

power shapes relationships between humans and nonhumans is useful for understanding the

dynamics of players and nonplayer characters. And this is especially the case when it comes to

antagonisms.

Because of the inherent differences between humans and computer game characters, and the

derivative, representational nature of computer game worlds, game enemies are both unique

entities and good examples of the pervasive logic of domination underlying Western systems

of thought. Game enemies are ethereal, electric creatures unlike any other being humans

interact with. But because they strive to resemble familiar situations, they end up re-shaping

old relationships and old roles into new and weird things. And so the logic of domination does

not work as straightforwardly in game worlds as in real worlds.

GENRE

FIGHTING GAMES

Entire games exist where fighting an enemy is the only activity the player can do. These games

(the series of Tekken, Mortal Kombat, Street Fighter, Dead or Alive, Fatal Fury, SoulCalibur etc.)

are like boxing matches in that there are two opponents, the player and the enemy, facing off

111

against each other. There is minimal plot, little environment, and the domination of one over

another is the entire point. The dynamic antagonistic interaction of two characters begins,

comprises, and ends the game in a stark example of what Juul calls an emergent narrative –

where a story arises from a player's individual experience in a game world to be recounted

after the fact, rather than determined by a writer before the player enters. While the beginning

and ending of the story are always the same (two buff angry people appear on a level surface;

one person dies and the other is chuffed), and the player is limited in what they can do

(generally punch, kick, block, dodge, attempt a combo move or quit the game), whether the

story of the game is one of a savage beatdown, an unexpected turnaround, an exhausting

struggle, or a noble give-and-take is up to the individual player.

Fighting games were originally made with two human players in mind, but it is also possible to

play these games against the computer, where the game controls the enemy. In these games,

the enemy provides a real challenge. The game AI makes instant decisions, always performs

complex combo moves perfectly, and takes no pity on the player. Fighting games, in fact, offer a

lovely image of antagonistic human-enemy togetherness as both characters are autonomous

volunteers, have the same aims and abilities, and their relationship is central rather than

tangential to the game experience.

Fighting games are about brief, intense relationships between two characters. The

relationship is straightforward and unambiguous, and the power dynamic is explicit and

involves evident to-and-fro between characters. The relationship is zero-sum and there are

only two outcomes: winning or losing. The relationship is not instrumental, but its existence,

its subtleties, and its development are the entire point of the game. In a way, fighting games

are akin to virtual pet or dating games in that the player’s attention is focused on a single,

significant other. But unlike (most of) those games, the relationship between a player and

nonplayer character in fighting games is combative and competitive. If the player makes a

wrong turn in a dating or pet game, it is the pet or paramour who is damaged, and the

relationship is soured. In a fighting game, if the player makes a wrong move, it is their own

player-character who is injured, and the relationship remains the same: hostile. While the aim

of the love game is to make the player feel like a significant other, the aim of the fighting game

is to make the player feel like a significant threat, not a supplicant or a patron but an equal.

112

But in all these games the player must prove themselves to be worthy. While a key similarity

between these games of intense relationships is subtle reactivity, a key difference is speed.

In a fighting game, the player must make decisions quickly. If they want to win, they must be

constantly aware of the other character, where they are, what they’re doing, their special

moves, their health level (though indiscriminate button-mashing works pretty well up to a

level.) Interaction with the other is the point of the game; there is no quest to get back to, no

puzzle the enemy has interrupted, no new area to explore, nothing better to look forward to.

The fight, and the promise of more serially monogamous fighting, is all there is. So the player

must move fast in reaction to the enemy, must make snap judgements on the best way to get

past the character’s defences and into their most vulnerable spots. The player must know

their enemy in order to hurt them more effectively, and they must also know their own

character’s special abilities.

Starting with 1991’s Street Fighter II, players of fighting games could choose their own

character from a group of fighters, the rest of which they must then defeat. Characters also

possessed their own special moves, like Dhalsim’s fire-breathing and limb-stretching or Chun-

Li’s speed and backflips. The special moves of characters tend to counter-act each other, so

one character’s leaping ability can be undone with other’s grappling hook. When the player of

a fighting game moves, they control distinct parts of their character’s body with set buttons:

the arms to punch or block, the legs to kick or jump. (From this often awkward mechanic has

emerged the fumblecore genre, featuring games like QWOP or Regular Human Basketball,

which challenge the player to move ragdoll characters limb-by-limb). Combo moves are

unlocked by quickly pressing a series of buttons and super-combos are even more difficult

combinations; E. Honda’s super-combo in Street Fighter II, for example, involves holding down

the left button, then quickly pressing the right button, then the left and right button, then the

punch button (this instigates the “Super Killer Head Ram”).

In a fighting game, the player and the enemy build an affect cycle between them. But this

reciprocal, interdependent loop is built on opposition rather than mutuality. Each must

protect their body and damage the other’s. They must affect without being affected. In

contrast to the affect cycle of the friendly game, in which the player and the character each

113

finds pleasure in pleasurably affecting the other, the fighting game sets the player and the

character as opposites. Instead of gaining pleasure by giving pleasure, the player and the

nonplayer must strive to affect each other negatively in order to affect themselves positively.

What the player wants, the enemy despises, and vice versa. The enemy’s pain is the player’s

pleasure, and the player’s pain is doubled by knowing their enemy benefits from the player’s

injury. While an owner of a virtual pet can build positive affect together with their character,

the player of a fighting game seizes pleasure from their counterpart and has it seized from

them in turn. In his discussion of enemies (or objects of hatred), Spinoza notes that “he who

conceives, that an object of his hatred is painfully affected, will feel pleasure … greater or less,

according as its contrary is greater or less in the object of hatred” (146). But it is not always

the case that the enemy of the fighting game is hateful. The enemy can be an opposite more

like a mirror image, a medium through which the player can affect themselves.

One gets the sense in fighting games that opponents have their own lives outside the game,

that they have friends, training regimes, and some pride in their abilities. Even the early

arcade fighting games had a surprising amount of narrative behind each pixillated character.

The characters are not fighting for their lives, but like the player, are playing a friendly game.

In this, their narrative role (sport competitor), their in-game persona (as a sportsperson), and

their functional role (as an opponent in a sport) are in harmony. But there is a fixity to these

roles, an inflexibility to who these characters are. They too have a primary essence that can be

defined, categorised, essentialised, and so, overlooked. Good characterisation, for them, can

only be expressed through fighting well. Their actualisation comes about through action, but

these actions are determined by the game. The characters can only develop in certain

directions.

The “friendship” finishing move in Mortal Kombat II and 3 offers players an alternative to

gruesomely killing opponents to end a match. The move can be played when the player has

defeated the opponent and only needs to deliver a final blow to be declared the winner – a

checkmate point. The moves are silly – giving the opponent a present, blowing them a kiss, or

doing a dance – but give the player a chance to avoid violence, sidestep their role expectations,

and change their relationship with their opponent from attempted murderer to goofy buddy.

They change the affective tone of the game from serious to silly. But these moves can also be

114

part of the domination of the enemy, a move to humiliate them, an expression of their

contempt and the enemy’s triviality. (The YouTube channel Gamer CHI-CHA has a rather

wonderful series of videos of all characters’ finishing moves, including friendship fatalities, in

Mortal Kombat II and 3).

RACING GAMES

Racing games, or sports games, have enemies of a different, competitive kind. Nonplayer

enemies in Mario Kart DS can still damage player-characters by firing bombs in front or

behind them, cutting them off, or covering the road with banana skins, but they can also

overcome the player just by being better drivers. The characters’ bodies are still the site of

competition, but touch is not required for one to succeed over another. Instead, the quality of

the body is at stake, the speed of its reactions, the surety of its manipulation of the controller,

the accuracy of its perception of the path ahead, its evaluation of opportunity and risk. As

computers are quite good at these things, characters controlled by them could trounce human

players perfectly every time (Cechanowicz et al).

The main ability needed by racing nonplayer characters is path-finding, a rather complex bit of

programming to let a nonplayer character perceive and navigate an environment. Pathfinding,

as Botea et al. tell us, is one of the more complicated tasks for a game AI, as “[m]oves have to

be computed in real time, and often there are many mobile units to compute paths for. Paths

have to look sufficiently realistic to a user. Besides a more standard, single-agent pathfinding

search on a fully known, static map, games feature more complicated variations of the

problem, such as multi-agent pathfinding, adversarial pathfinding, dynamic changes in the

environment, heterogeneous terrains and mobile units, incomplete information, and

combinations of these.” (21). Simple games set nonplayer characters on predetermined paths

from which they cannot deviate or must immediately return to. These paths can be

interrupted by actions of the player-character, but the nonplayer character cannot decide

where to go on their own. Horizon Zero Dawn nods to older games by making these paths

115

overt, glowing brightly in the player’s heads-up display. Its robot dinosaurs are proud of their

rails. In racing games, pathfinding is at once easy, because the characters are just going around

and around a set track, and difficult, because characters constantly encounter new obstacles,

and must keep themselves within a sporting distance from the player (like other enemies,

racing enemies are made to be beatable.)

[Image removed due to copyright restrictions]

Fig. 9: In Horizon Zero Dawn, the player can see the predetermined paths of their enemies.

(author screenshot)

This has its limits, however. “Dynamic difficulty adjustment (DDA)”, as Compton explains, “is a

process in which video games will adjust the difficulty of the game over time based on the

player’s performance. If the player is doing well the game will become more difficult to add

challenge. On the other hand, if the player is struggling the game might reduce the difficulty to

keep them from getting stuck”. In terms of pathfinding, this produces the “rubber band effect”,

where nonplayer characters bounce around the player-character as if they were connected by

elastic. “When the player is ahead the opponents speed up, and when the player is behind they

slow down. The further ahead or behind the player is, the more the computer opponents

react” Compton clarifies. Dynamic difficulty adjustment means that characters are affected by

the player’s performance, even when the player isn’t attending to them. Though most obvious

in racing games, DDA can be found in all sorts of situations where things need to change to

accommodate a particularly lucky or unlucky player.

In 1974’s Speed Race, enemy cars act less like characters at all, and more like environmental

obstacles around which the player has to dodge and weave. The aim of the game is to keep

driving uninterrupted for as long as possible along a long straight road, which is made difficult

as enemy cars keep trying to crash into the player-character. As game design gradually moved

away from getting the player to pop another 20 cents in the arcade machine, and realism

became something to strive for, the role of enemies in racing games became less about game-

ending destruction and more about legitimate competition and the joys of overcoming odds.

In go-kart-style racing games like Mario Kart DS, players race each other, but can also engage

116

in high-speed combat to hobble other characters. The basic mechanics of modern racing

games are pretty much the same whether nonplayer characters exist in them or not.

Opponents exist for the purpose of comparison and competition, a little bit of extra

motivation, someone who the player can assert superiority over. But these games can also be

played perfectly well if the player is trying to beat their own personal best or some arbitrary

number. Computer-controlled enemies in racing games, then, facilitate relationships between

the player and themselves.

Pacheco et al’s study on the believability of nonplayer characters in racing games illustrates

the difficulties in creating affective competitive game characters. They note that “Games seem

to be more engaging to players when they believe they are facing/cooperating with other

human players” (1), and subsequently suggest that making NPCs more human-like will make

them more engaging. But as they concede, player recognition of humanness is something quite

difficult to test for (1). In Pacheco et al’s turing-test-like racing-game environment, players

identified other cars as being controlled by the computer when the character drove well,

drove badly, seemed overcautious, seemed overconfident, took every opportunity, didn’t take

opportunities, drove too slowly and drove too quickly (8). They also didn’t enjoy computer-

controlled or human-controlled nonplayer characters any more or less than each other.

Pacheco et al. conclude that players’ evaluations of nonplayer characters are highly contextual

and players have different needs for different types of games. In racing games, players didn’t

have a strong need for human-like characters, but instead wanted difficult competitors that

would provide a challenge; an enemy that could match them in skill. Pacheco et al. recommend

that “the agent should behave according to the situation it is in … Introducing believability is a

mechanic that has to be thoroughly considered … doing so might interfere with the enjoyment

of a particular game” (9).

STEALTH GAMES

117

In stealth games, enemies must be avoided rather than confronted. They invite the player to

overcome from underneath. In Invisible, Inc. players must creep around an office building at

night, avoiding and incapacitating guards without being detected. Invisible Inc. is an isometric

third-person turn-based game in which the player must gather information to sneak two

player-characters past guards in order to steal items. Players receive a series of missions to

steal something from someone, and must work out how to best complete it. Although the

game is isometric, the player can only see what is visible in the player-characters’ own vision

cones, which is in colour, and a remembrance of what the player-characters have previously

looked at, in grey-scale. At the beginning of a mission, this means that most of the game level is

invisible to the player, and the player must tentatively explore around without being caught.

The guards’ vision cones are also visible to the player, so the player is aware the direction in

which they are looking. Guards patrol on set paths, which the player see if they choose to

spend one of their limited action points. If the player-character is at the edge of a guard’s

vision cone, the guard will walk toward the player-character to investigate. If the guard looks

straight at the player-character, they will shoot them immediately. Guards are observant, and

constantly looking around. The player must imagine the game level from their point of view.

The game is a race to see who can gather and use information about the other first. Players are

outgunned and outnumbered, and if a player-character is seen all hell breaks loose. As a

mission continues, more guards appear and begin using other sensory devices like remote

scanners to detect the player-character.

Enemies can register a sneaking player to different degrees. Guard enemies, in particular, have

the ability to doubt themselves. When a guard has briefly registered the player-character in

their vision cone, or a hidden player-character has made a noise, a guard can stop their patrol

and become alert for a time. In this temporary state, they may actively search for the player-

character around a certain area, or attack the player-character in their hiding spot if the

player-character makes another move or sound. But if the player-character stays quiet and

still, the guard will, after a certain time, relax and return to ignorance. Like adults in a child’s

game of hide and seek, the nonplayer characters do know where the player-character actually

is (because they are connected to the omnipresent game world) and are only pretending they

don’t. In Horizon Zero Dawn, guard dinosaurs, shown below, glow blue when they have not

detected the player, yellow when they have detected the player and are alert, and red when

118

they are attacking. This traffic-light system informs the player of the enemy’s internal state,

and lets them predict their behaviour. The player has to put themselves into the robot

dinosaur’s shoes in order to figure out how to best act. The player and the enemy, then, find

themselves in a rather tense affect cycle. While the enemy cycles through three states

prompted by temporal and sensory triggers, the player inhabits a variety of states in dialogue

with the nonplayer character.

[Image removed due to copyright restrictions]

Fig. 10: A happy Watcher. (author screenshot)

[Image removed due to copyright restrictions]

Fig. 11: An alert Watcher. (author screenshot)

[Image removed due to copyright restrictions]

Fig. 12: Three attacking Watchers. (author screenshot)

Familiarity with an area can allow the player to avoid an enemy, or at least to know the point

where the enemy will stop chasing and turn back. In stealth games, studying an enemy is key

to predicting their behaviour and using it as a reference for future action. The point of a

stealth game is to avoid affecting an enemy at all. A happy stealth game enemy is one that

never encounters a player, and vice versa. But sometimes enemies are unavoidable. When

guards are protecting an area, particular tactics are required. To defeat guards requires

sneaking, hiding, long-range weaponry and awareness of surroundings. For although they may

be individually weak, in a group guards can easily overwhelm a single player-character. So the

player must find nooks in which to hide, and be patient, anxious, alert and opportunistic. They

must be aware of the relative locations of the guards, and the likely direction of their

patrolling. They must have a sense of the guards’ line of sight. The player must be aware of the

perception of the enemy character, how far it can see, and whether obstacles block their view.

And so, in waiting to kill the guards, the player takes on their perspective, to a limited extent.

119

ENEMY AFFECTS AND EFFECTS

The ways in which humans and enemies affect each other, then, are not unique, but they are

particular. They offer specific pleasures and challenges in specific ways. Through a rather long

look at the things humans and enemies can do to affect each other and form relationships, this

section will attempt to tease out the mechanisms of human-enemy affect.

Enemies act and are acted upon through particular mechanisms. Some of these mechanisms

are unique to enemies, and some are repurposed from other areas of the game. Game enemies

look very different to each other, but they all have some basic behaviours in common. They

might look spiky or slimy, they might go for the throat or shoot bullets from afar, they might

move fast or stand their ground, but they are united in three fundamental things: they are

spawned, they attack the player, and they die. In doing these things they are inexorable; there

is nothing so reliable as a computer game enemy. In the tradition of what Brown and Stenner

call “the technique of unsettling familiar ways of thinking by contrast with the radical

otherness of exotic ways of thinking [and which is] by now a familiar strategy in the repertoire

of social constructionism” (81) and what Plumwood (2007) identifies as “that recent radical

tendency in environmental philosophy that has stressed that respect and attention must go

‘all the way down’ and need have no boundary or limitation” (19), this section will attempt to

describe what affect is to an enemy nonplayer character. This will be examined first through

what enemies do to themselves, and then what players do to them.

SHOWDOWN

ENEMY SPAWNING

120

When an enemy spawns, it modulates from an entity of code (that may or may not already be

running some functions) to a character in an environment that can move, act, and be seen. It

has, from the viewpoint of the player, just popped into existence. However, the enemy is a

creature with no childhood. It arrives fully-formed, a robot that’s been instantly assembled

according to an existing blueprint. This manufacturing of the enemy is always triggered by

something; to save processing power, games do not spawn all their enemies all at once. In

different games the trigger may be the loading of a new game area, a certain time after some

event, when the player interacts with a certain object, a change in health or level of other

enemies, the death of other enemies, the spawning of companion enemies, a change in health

or level of the player character, or the proximity of the player character to a spawn point (a

specific place within a game level at which enemies have been programmed to appear).

To spawn does not necessarily mean to start existing, but it does mean an increase in the

enemy’s ability to affect and be affected. An enemy’s code may be monitoring the location of

the player and recording some of their attributes in order to determine when and where to

render its sprite or model in the game world. In this sense they have some limited ability to

act. Once an enemy is spawned, they have many more behaviours available to them – in

Spinoza’s words, they have become more powerful, and for a human this would result in the

conscious experience of pleasure. But functions of enemies are usually not so concentrated

within a single identifiable object, and enemies, before they are spawned, can be said to be

distributed across many game functions if they can be said to exist at all. A character like

GlaDOS of Portal, who is invisible for most of the game, is made up of voice files, timed

environmental events, and only acquires a visible body at the end of the game. Her ‘spawning’

does not result in an increase in her power, because her character acted through the game’s

level design. Her body, if she had one, was made of architecture and sound. The idea of her as

an individual only exists in the mind of the player.

When and where enemies spawn can create different experiences for the player. Some

enemies, like the Colossi of Shadow of the Colossus, are threatening, faraway presences that the

player can anticipate fighting. Some enemies silently spawn close by and take the player by

surprise. The carnivorous lizards of Rain World spawn in nearby tunnels, resulting in an

apprehensive, jumpy player experience. The robotic shepherds of Horizon Zero Dawn and the

121

soldiers of the Metal Gear series spawn long before the player arrives in an area and are keen-

eyed guardians whose mere gaze must be avoided, resulting in a sneaky, wary player

experience. The wolves and rats of Skyrim have spawn points just randomly scattered around

the countryside, triggered by player proximity, and can be fought or fled from depending on

one's disposition, resulting in an opportunistic player experience.

The long-running Final Fantasy series is a good demonstration of how enemy spawning has

evolved. Many older, simpler games have also been ROM-hacked, meaning their code has been

reverse-engineered and is publicly available (the legality of this is under debate). This

information is used by speedrunners to figure out how to best avoid enemies, and so complete

the game faster. It also gives us an insight into how enemies are programmed, and the

decisions designers made behind the scenes.

In early games of the Final Fantasy series, players travel through wilderness and towns,

encountering enemies in random locations and bosses in predetermined locations.

Randomised enemies make moving through the world seem exciting and dangerous, and make

the games more interesting to replay and compare with friends. As the early Final Fantasy

games often involved a lot of backtracking through areas, enemies appearing in unexpected

spots also made gameplay less predictable.

In 1987’s Final Fantasy (FF), the limitations of the NES console meant the appearance of

randomisation was tricky to achieve. Enemies in FF surprise the player and whisk them away

to a separate battle screen based on the number of steps the player character has walked (see

AstralEsper’s FF game mechanics guide for a thorough catalogue of the step numbers at which

enemies will appear). Speedrunners use this information to count their steps and make sure

that they take their critical enemy-spawning steps somewhere where enemies are not allowed

to spawn, like an indoor area. The step counter also resets when the player enters a different

area. The types of enemies, the amount of damage they can do, the amount of damage the

player can do, and the player’s fighting method are determined by random number

generators. Persistent memory is stored by a battery in the game cartridge itself, so when

players fire up their NES and start a new game of Final Fantasy, this information is stored by

the game and not the console. The locations of enemies are set until the player finishes the

122

whole game and watches all the credits until the end (or the cartridge glitches or is dropped),

but if the player starts a new game after that, the enemies have different attributes.

Final Fantasy IV (FFIV) also spawns enemies in ‘random’ locations. The ‘random’ locations of

enemies is actually determined by the number of the frame of the title screen when the player

presses ‘Start’, and are based on the number of steps the player character subsequently moves.

Through this, the player inadvertently sets their own enemy spawn points. FFIV runs at 30

frames per second. If a keen gamer presses ‘start’ exactly two seconds into the title screen,

then the number ‘60’ is the ‘seed’ for a set of pseudo-random numbers (numbers that seem

random to the player but have been chosen by someone at some point). These numbers (for

example: 18, 35, 45 and 90) mean that the player will encounter enemies after they have made

18, 35, 45, and 90 steps. (However, to make things tricky, steps on certain surfaces, like rugs,

stairs, or alongside furniture, are not counted as steps.) Speedrunners take advantage of this

setup by deliberately starting the game on a certain frame, which they know will seed certain

pseudo-random numbers. They can then make sure that their 18th, 35th, 45th and 90th steps are

taken in places that enemies cannot spawn in, like friendly pubs or doorways. Because this

information was stored on the SNES console, rather than the cartridge like Final Fantasy, the

locations of enemies were not determined for the whole game, but reset whenever the player

restarted a saved game. The first way players affect enemies in these games, then, is to

unwittingly place them. Final Fantasy VII (FFVII) also uses a step count to place enemies,

though its walking and running systems count steps differently. Final Fantasy games after FFVI

based player position on the player character’s velocity rather than a landscape’s grid, making

step counting an inefficient method for enemy spawning.

Final Fantasy XII (FFXII) is the first Final Fantasy game where players can see an enemy ahead

of them and walk up to them, rather than be surprised and taken to a special battle scene. In

Final Fantasy XIII (FFXIII), enemies always spawn in the same place, no matter how long the

player dawdles on the title screen. FFXIII is a more linear game than its predecessors; players

are not expected to backtrack or revisit places (so much so it has been called a “corridor

simulator” (Treharne)). Because players only visit an area once in a game, enemies do not

need to reappear in random locations. Interestingly, enemy spawning in FFXIII is triggered by

the player’s camera, and enemy attacking is triggered by the player-character’s field of vision.

123

FFXIII uses a 3rd-person camera which the player can control. If the enemy’s spawn point is in

the camera’s field of view, then the enemy will render. But if the player points the camera at

the ground, the enemy never spawns, and the player can just walk through the level

unchallenged. The enemy’s sprites show up in the player’s HUD’s mini-map, but the enemy’s

sprite and AI does not activate until the player can actually see them. Rendered enemies in

FFXIII attack the player on sight – that is, they have a sense of sight, and so a line of sight. If the

player is behind an enemy, the enemy cannot see the player, and so does not attack them. In

these cases, the player can perform a sneak attack for extra damage. If the player never lets

the enemy render (by walking around with the camera pointed to the ground), then the

enemy’s vision cone is also not rendered, and it cannot perceive or attack the player. If the

player-character runs around looking at the ground, the enemy doesn’t attack. The enemy

doesn’t exist unless the player is looking at it; the enemy does not acknowledge the player

unless the player looks at it. In this case, the childhood conceit of covering your eyes so the

monsters won’t see you actually works.

PLAYER RECOGNISING

“Often,” Eco states, “when faced with an unknown phenomenon, we react by approximation;

we seek that scrap of content, already present in our encyclopedia which for better or worse

seems to account for the new fact” (57). Isbister (30), Schell (255) and Cook argue that

enemies must be instantly recognisable so the player can immediately act against them. So the

first thing that human players do towards enemy characters is judge them. Enemy characters

have a programmed certainty about the nature of other characters, but players must learn and

form associations from contextual and intertextual clues.

Associations allow players to move quickly against characters that can be instantly judged to

be one type or another based on appearance and pattern of movement. The need for players

to recognise enemies as enemies results in a stratification of classes and types of enemies.

Enemy types then need to be as differentiated as possible. This is why beings in games are

characters; they have characteristics.

124

Enemies occur in types, and games usually contain a few different types of enemies.

Characterisation occurs on the level of species, not individuals. Within a game, enemies are

visually and behaviourally homogeneous within a type but dissimilar between them. This

categorical difference is evident visually and behaviourally. Because “clear silhouettes

distinguish different characters from one another” (Cooper), designing enemies with

distinctive shapes and animations helps the player recognise and categorise them. Enemy

bodies may all be variations on a theme, like Horizon Zero Dawn’s robot dinosaurs or Rain

World’s lizards, but their type is recognisable from a distance and their actions are unique to

their kind. Their commonality is in their antagonism towards the player. A player can

encounter a Glinthawk in Horizon Zero Dawn once, learn its range of actions, strengths,, and

weaknesses, and then successfully apply that knowledge to every Glinthawk encountered

thereafter, because all Glinthawks look and act the same. Enemies operate from extremes of

sameness and difference.

This type-casting of enemies provides the thrills of predictability. Because game characters do

not have good memories (Kasap et al) and do not form associations, they cannot learn as well

as players do. To keep play challenging, games simply add new enemies over time, so what

enemies lack in diversity of behaviour they make up in diversity of types. Because they are

novel, new enemies are a challenging encounter. They also give the sense that every man and

his dog is out to get the player – a feeling integrated into the narrative of post-apocalyptic and

anti-hero games. When a player encounters an enemy they recognise, the player can respond

with confidence, but when they encounter an unfamiliar enemy they must be more cautious.

This change in enemy quality and quantity is important for a player’s sense of progress. As

well as new armour, weapons, skills, costumes, areas to explore or progress bars, the player

can measure their success with the type of enemy they encounter. As a game proceeds, game

enemies change from small and piddly to large and mighty, culminating in a boss battle with a

particularly difficult and unique enemy. The Elder Scrolls series epitomises this tendency. In

The Elder Scrolls IV: Oblivion the player begins their game being attacked by delicate rats and

slow mud crabs. As they explore more of the game world and level up their skill, the game

introduces new creatures, like wolves and pumas, and eventually more dangerous and

125

humanoid monsters. These enemies are tied to location and the player’s level; the player can

only encounter them if they are in a certain place and have achieved a certain status. By the

end of the game the player is attacked by gods and demons – beings that in the narrative of the

game did not initially recognise the player as a threat, and in the mechanics of the game did

not even exist until the player had become very good at killing.

The requirement for enemies to have a visually identifiable essence can cause problems if

games are trying to represent reality. Representation doesn’t have to be very accurate in order

to be affective, as long as the human player is willing. As Eco states, “Does a 3-D model of a

man correspond to a concept of ‘man’? Certainly not as the classic definition (mortal rational

animal) is concerned; but as far as the possibility of recognising a human being is concerned,

and then of being able to add the determinations that derive from this identification, it

certainly does” (86). In other words, people get the idea and go along with it. However, quick

categorisation and essentialisation of others does those others a disservice. Survival games

like The Long Dark are quick to point out that wolves in reality are pretty chill and you

shouldn’t shoot them on sight.

[Image removed due to copyright restrictions]

Fig. 13: A disclaimer from the The Long Dark asks players not to shoot real wolves. (author

screenshot)

Objectifying others ends up objectifying the self in an opposite role, as Foucault (277) and

Plumwood (31) point out. To identify an essence in another is usually an act of creation rather

than discovery, but as nonplayer characters are designed objects with prescribed purposes,

this complicates things. von UexkuH ll calls this dual construction of the self and designed other

‘counterpoints’. He states that “A coffee cup with a handle shows immediately the contrapuntal

relation to coffee, on the one hand, and to the human hand, on the other” (191) (though surely

the point of cups is so you don’t get coffee on either of your hands). Game characters likewise

find their form through the needs of the game and the needs of players. von UexkuH ll further

notes that “The meaning of our use-object for us lies in its function, and this function can

always be brought back to a bridging of the counterpoint in the object to the human being,

126

which at the same time forms a motive of the bridging” (191). Enemies are the way they are

because their function is to make the player powerful. In games, this produces a particular

process of rationality, as Foucault would put it, or another iteration of mastery, as Plumwood

would. Recognition, or the categorisation of others into types, is the beginning of the

immediate human-NPC power dynamic.

FIGHTING

ENEMY REACTION

Enemy fights are exciting and rewarding, and programming them is relatively easy. Isbister

argues that enemies are both exciting and simple because a human's perception of another's

unfriendliness is both instantaneous and psychologically compelling (228), but there are

other reasons for enemies' across-game ubiquity and in-game similarity: convention, low

player expectations, and variables that count down. Enemies are expected to exist, and they

are expected to express their feelings through action. Popular game engines, such as Unity and

GameMaker, contain complete programming templates for creating enemies, and tutorials for

game design beginners usually have students designing simple combat systems. In this way,

enemies are instituted as essential but simple game elements. The ease of making characters

move towards or away from the player, or to appear when the player is in a certain location,

forms the basis for quick enemy development. One needs only to make the enemy character

move toward the player and tell the player they are losing something (whether health, gold

rings or time) when their character is touched by the enemy to make an affective interaction.

Interaction with an enemy follows a pleasing three-act structure. An enemy and player

character come across each other, they fight, and one of them dies.

Treanor et al. (in ‘AI-Based Game Design Patterns’) observe that “the most common role for AI

in a game is controlling the non-player characters (NPCs), usually adversaries to the player

character” (1). All enemies are controlled by some sort of artificial intelligence (AI), though

127

the intelligence part is a bit of a misnomer. AI in games refers to various programming

techniques of simulating intelligence in nonplayer characters. It is an architecture, a way of

structuring code that affords decision-making and smooth transition between behaviours. The

most common ways of structuring code in games are behaviour trees (where characters move

down a tree of behaviours according to conditions (Simpson)), finite state machines (FSM)

(where a character continues in a state of being until it’s told to stop (Rasmussen)), planners

(where the character has a goal and selects from a set of behaviours those that will help it

reach its goal (Mark)) and utility (where potential actions are scored for their suitability to

achieve a goal (Mark)). These architectures are different ways of structuring the same

substance (of behaviours and transition between behaviours), and aren’t really visible to the

player (Mark), but are appropriate for different types of character behaviour. Orkin

summarises the situation in remarking that “the reality is that all A.I. ever do is move around

and play animations! Think about it. An A.I. going for cover is just moving to some position,

and then playing a duck or lean animation. An A.I. attacking just loops a firing animation.

Sure there are some implementation details; we assume the animation system has key frames

which may have embedded messages that tell the audio system to play a footstep sound, or the

weapon system to start and stop firing, but as far as the A.I.’s decision-making is concerned, he

is just moving around or playing an animation” (2).

However, the type of AI architecture used in a game has some bearing on how characters are

affected, and by what. It is not necessary to go into too much detail – in terms of affect AI

programming is to NPCs what cell physiology is to humans or petrology is to rocks. Every

game will have their own unique variables that modify and trigger changes in characters,

much as Spinoza says that the affect of one individual differs from another according to their

essences (169). As an example, let’s say an NPC walks up to a locked door. This NPC will be

affected by the locked door whether it is governed by a behaviour tree, a FSM, or a planner, but

this affect will be processed differently according to each one. For ease of reading, let’s call the

behaviour tree NPC, FSM NPC, planner NPC and utility NPC Blossom, Bubbles, Buttercup and

Bingo respectively.

128

When Blossom comes across the locked door, she has already been continually climbing and

falling out of her behaviour tree once per ‘tick’ (a small predefined unit of time). Getting to the

door instigates a sub-tree of behaviour that looks a bit like this:

[Image removed due to copyright restrictions]

Fig. 14: An abstracted model of a behaviour tree for opening a locked door. (author supplied)

Blossom will go through this behaviour tree left-to-right and top-to-down. If a behaviour is not

possible she will go back to the last possible behaviour. If she cannot open the door she will

search her inventory for the key and if she does not have the key she will try to pick the lock

and if she succeeds in picking the lock she will open the door and if she opens the door she

will walk through the door and close it behind her.

Bubbles probably doesn’t encounter a locked door, because finite state machines probably

wouldn’t be able to handle that complicated a situation. If she did, she would change from her

walking state to a predefined unlocking-the-door state, and if that was successful she would

once again change to a walking state and go through the door.

When Buttercup happens across the locked door, she possibly runs away from it to find an

easier way around. If getting through the locked door is the only way to get to where she

wants to go, she will observe that she she has no key in her inventory and her lockpicking skill

is low, so she might decide to fetch a key if she knows where one can be found, search for a key

if she doesn’t know where to find one, or practice lockpicking on the door to increase her skill.

Buttercup likes using the least amount of steps to fix her problems.

When Bingo eventually reaches the locked door, she weighs up the pros and cons of each

possible approach. All of her potential actions have been weighted ahead of time, but have also

been modified during gameplay. Bingo considers finding a different path, finding a key and

learning lockpicking, and discovers that blowing up the door with her grenade is her highest

rated potential action. Bingo prefers certain methods over others, no matter how long they

take.

129

AI research likes to talk about sophistication, complexity and believability, and tends to use

these as synonyms. These terms refer to the ability of characters to perform a larger number

of behaviours, particularly sequences of behaviours, especially context-aware sequences of

behaviour, that a human can find understandable and justifiable. Good AI creates characters

that are convincing (Simpson), immersive, deep and life-like (Rasmussen), and intense

(Orkin). Bad AI results in silly characters that make unrealistic decisions and break

immersion. Better AI makes NPCs more human-like, though Magerko et al. point out that “To

date, every successful computer game is an existence proof that you can create good games

without human-level AI” (877).

One game that definitely doesn’t have human-level AI is 1994’s Sonic the Hedgehog 3 and

Knuckles. Although enemies of every game genre behave differently, the simple enemies of the

platformer demonstrates their quintessential qualities. This is not to say that platformer

enemies do not have their own peculiarities. In platformer games, the player moves fast,

jumping from one platform to another. Enemies in these games are more akin to physical

obstacles than characters, animated objects that get in the way of the player. The main

relationship in the platformer is not between the player and other character, but the player

and themselves, as mediated through their own performance. Platformers promise freedom of

movement, and frustration of that freedom of movement. Platformers, then, exemplify

enemies as impediments to the player’s relationship to themselves and their goals.

PENGUINATORS

In Sonic the Hedgehog 3 and Knuckles, a 2D sidescrolling platformer, the player encounters

badniks, animal-shaped robot suits powered by enslaved woodland creatures. When the

player defeats one of these enemies, the robot shell splits open and these cute critters are

freed. Having already played Sonic 1 and 2, the player knows that these robots have been

created by Dr Robotnik in his quest to mechanise the world and create an alternate ecology of

machines. The enemies of Sonic the Hedgehog 3 and Knuckles come in a variety of shapes, and

130

their attacking behaviour is tied to their form. A robot monkey, for example, throws projectiles

at the player, and robot piranhas bite the player character and drag him down into the water.

In this game, the player can play as Sonic the hedgehog, Tails the fox, or the correct option,

Knuckles the echidna. Two players can play together as Sonic and Tails, or Tails can

accompany the single player as an AI-controlled companion. These three player characters

have slightly different ways of playing. Sonic can run fast and put up a shield. Tails can fly and

swim. Knuckles can’t jump as high as the other two characters, but can fly and punch.

Knuckles, who was designed as a kind of hard mode, faces different and more difficult enemies

than the other two. Being spiky, all player characters damage enemies by curling into balls and

jumping or rolling into them.

Most enemies in Sonic the Hedgehog 3 and Knuckles only need to be hit once to die, though a

few are armoured and take two or three hits. Enemies in this game utilise hurtboxes and

hitboxes. A hurtbox is an area of a character’s body that can hurt others, but does not

necessarily register any hits itself. A hitbox is an area of a character’s body that registers hits.

As Gladstein puts it, “A Hitbox is an invisible box (or sphere) that determines where an attack

hits. A Hurtbox on the other side is also an invisible box (or sphere), but it determines where a

player or object can be hit by a Hitbox … [hurtboxes] can register a hit, but it should not collide

in the physical sense. Hurtboxes should only interact with Hitboxes. [Hitboxes] should only

interact with Hurtboxes.”

[Image removed due to copyright restrictions]

Fig. 15: The hitbox of a Buzz Bomber. (Lapper 2 ‘BuzzBomber’)

[Image removed due to copyright restrictions]

Fig. 16: The hitbox (yellow) and hurtbox (blue) of a Crabmeat. (Lapper 2 ‘Crabmeat’)

[Image removed due to copyright restrictions]

131

Fig. 17: The hitbox (pink) and hurtboxes (red) of a Caterkiller. (Lapper 2 ‘Caterkiller’)

Figures 15 to 17 show the hitboxes of three characters from Sonic 1. (The same system was

used in Sonic the Hedgehog 3 and Knuckles). Figure 15 shows that the Buzz Bomber’s hitbox

and hurtbox are the same, and extend outside the boundaries of the visible character. If the

player-character touches that blue rectangle, they will die. Figure 16 shows that the Crabmeat

character has two hitboxes: a blue one that registers the player-character and a yellow one

that registers the environment. When Crabmeat walks over to the rock on his left, its claw will

appear to pass in front of the rock, before the yellow hitbox touches the boulder and Crabmeat

is redirected. Crabmeat can sense the player-character with a larger part of its body than it

sense its surroundings. But if the player hits Crabmeat on the tips of its claws, Crabmeat will

not feel it. Figure 17 shows the complicated arrangement of a Caterkiller. The Caterkiller can

only receive blows on its head and can only inflict blows with its three body segments. Hitting

the body does not damage the Caterkiller, and being hit by the head does not damage the

player-character.

So when the player, playing as Knuckles, comes across an enemy Penguinator in the IceCap

Zone of Sonic the Hedgehog 3 and Knuckles, they can jump in an attempt to avoid it, or jump or

glide in an attempt to hit it. The Penguinator knows to waddle back and forth within a specific

area and knows the location of Knuckles. If Knuckles gets within a certain distance of the

Penguinator, the Penguinator is triggered into a new behaviour. It will start tobogganing. The

Penguinator will still waddle forward, but its animation will speed up, as if it’s taking a run-up.

Its upright, walking sprite will be replaced with a horizontal, sliding sprite. It will move at a

faster, constant speed in whatever direction it’s already facing (which may be away from the

player). It will continue to move back and forth within its bounding box, but sliding fast

instead of walking slowly. At the edges of its bounding box, it will come to a complete stop. It

will stand up and turn around – changing its sliding sprite to its walking sprite, flipping its

walking sprite around, changing once again to a (reversed) sliding sprite, and sliding in the

other direction. The Penguinator continues doing this as long as it is visible on the screen, but

if the player leaves the area and returns it is reset to waddling. Because Knuckles moves so

fast in comparison and travels through areas quickly, this process is not often visible to the

player, who might catch the beginning of one sliding cycle at best.

132

[Image removed due to copyright restrictions]

Fig. 18: A Penguinator attacks. (author screenshot)

The Penguinator hits Knuckles by touching him when he is not jumping or dashing. If the

Penguinator hits Knuckles, the Penguinator pushes Knuckles along with it for a few frames,

the game checks that Knuckles has more than zero gold rings, and Knuckles loses his whole

gold ring collection and flashes invisible for one second. He is temporarily invincible and

cannot be hit again for three seconds. The Penguinator, at the end of its sliding function,

performs its standing-up-and-turning-around function, then slides in the opposite direction. If

this is within Knuckles’ 3 seconds of invincibility, it will slide past Knuckles without either of

them sensing each other. Otherwise, the Penguinator will register the touch, the game will

check that Knuckles has zero gold rings, and the game will instigate the end of the game.

If Knuckles successfully hits the Penguinator, either by jumping on it or punching it, it

registers the hit and instigates its own death. Its sprite is replaced with a sprite of a broken

robot, and a randomised seal ‘Rocky’ or penguin ‘Pecky’ sprite pops out and slides quickly

along the platform and off the screen. This sprite recognises no obstacles, only gravity and the

level architecture. If Knuckles unsuccessfully hits the Penguinator, Knuckles receives damage,

as above.

The Penguinators are slow, silly characters. Their bug-eyed, gormless expression and

awkward walking style, indicates to the player that they are not a major threat. They take a

second to warm up before performing their tobogganing attack (a long time in a Sonic game),

and they often slide in the wrong direction anyway. The player can easily avoid them, because

the player character moves so fast and the Penguinators react so slowly. But their essential

qualities of having a defined area to patrol, having an idle function and an attacking function,

being able to kill and be killed, having a fundamental antipathy to the player, reacting to the

player’s presence and touch, and looking different after death, are fundamental enemy

qualities.

133

Much AI research has focused on identifying and implementing specific behaviours that

researchers think will make characters more affective. Megerko et al. summarise this by

saying “there are inherent limits in the complexity of the behaviors required to create

compelling bots that are essentially computerized punching bags. Furthermore, these types of

games limit the human gaming experience to violent interactions with other humans and bots

… [to have any other kind of interaction] we really need complex AI characters” (877). A

general trend though, has been to let characters be affected by more things. More complicated

AI allows nonplayer characters to be affected by more things, more often. This in turn results

in more complicated behaviour. The trouble with allowing nonplayer characters to perceive

and react to so many variables is that it becomes hard to keep track of and debug

(Rasmussen).

HORIZON ZERO DAWN

A more recent example of AI is in the game Horizon Zero Dawn (HZD). The robot dinosaurs of

HZD interact with each other, form groups of mixed types, and flee from the player as well as

fight. In the narrative of the game they are all separate robot bodies controlled by a central

hivemind, but that’s sort of how they work as computer game elements as well. Thompson

(‘part 1’) explains that while the dinosaurs do function as individuals, invisible layers of AI

systems allow these individuals to function as collectives. There is, in essence, a visible

individual dinosaur, and above that an invisible ‘herd’ character who orders the individuals

around, and above that an invisible ‘collective’ character who orders the herds around. “A

group agent doesn’t physically exist in the game world, but it’s responsible for coordinating all

agents within it. Each individual agent has its own set of sensors that enables it to hear noises

close by or spot threats in proximity. But in addition, a group agent has a blackboard system

that stores in-game information a machine might want to use. This not only allows the game to

keep performance costs down by storing information each individual might try to calculate on

its own – like safe areas to stand in or active patrol routes – but more importantly, agents in a

group can share information between each other” Thompson explains. A single Grazer is only

the visible gestalt of a hierarchy of AI characters. This system is handed by a hierarchical task

134

network planner AI, meaning that characters are given goals, have a range of actions they can

choose from to achieve those goals, but the programmers have weighted those actions so the

characters will tend to prefer some over others. Each character in this chain – the visible robot

dinosaur, the invisible herd character and the invisible collective character – are given tasks by

those above them, and use the collective information stored by them in order to perceive their

world and choose between actions.

Robot dinosaurs in HZD prefer to be in groups and like to have something to do. When an

individual dinosaur is without a herd (if, for example, the player has killed all of their friends),

it asks the collective to assign it to a new one. To assist their transition from one group to

another, “each machine has a passport!” Thompson enthuses, “Their passport stores facts

about that machine, such as its level and machine type and the collective uses this to

determine whether it will fit the requirements of another group, only moving [it] across if it

satisfies the groups needs”. When a robot dinosaur is accepted into a new group, it is given a

suitable job to do, depending on its species.

Despite their being part of groups, individual robot dinosaurs in HZD still do a lot for

themselves. Each type of robot dinosaur has a different way of sensing the world, from the

Watchers’ vision cone to the Longlegs proximity sensor, and these sensors have a greater or

lesser sensitivity to stimulus (Thompson ‘part 2’). Characters do not only gather information

for themselves but are given it by objects. Objects like the player-character, other human NPCs,

other machine NPCs, arrows, rocks and animals carry information packets that they give to

their characters (Thompson ‘part 2’). The packet tells the dinosaur information about the

object and its current state. If a robot dinosaur comes across a body, that body will tell the

dinosaur whether it is alive or dead, which will affect the dinosaur’s behaviour toward it. If a

robot dinosaur comes across the player-character hiding in long grass they will sense it, but

the information packet will tell the dinosaur that the player-character is currently invisible,

and so the dinosaur will act as if it hasn’t seen anything. Not all the information in information

packets is recognised by all robot dinosaurs, however. Thompson observes that “certain

information might be ignored by some characters whereas others react to it promptly. In fact,

depending on the strength of a sensor in a given machine, it can actually reduce the data from

the sensory event that they can read” (Thomson ‘part 2’). Unlike the Penguinators, who take

135

information from the world around them, the dinosaurs of HZD are also given information.

Their senses are thus de-individualised, their bodies diffused.

When it comes to fighting, all these processes add up to something pretty simple. If the player

attacks one robot dinosaur in a group, the herd character does not update that information

instantly, so the it takes a while for other robot dinosaurs to be informed. Once a robot

dinosaur is aware of the player, either by sensing them directly or being informed by the herd,

the attacking sorts of dinosaurs attack, and the herbivorous sorts of dinosaurs tend to run

away (unless they’re cornered or there aren’t enough aggressive dinosaurs around). The herd

also makes sure that robot dinosaurs take turns attacking and don’t rush the player-character

all at once, which makes for a more exciting battle for the player. Smooth animation presents a

challenge, since robot dinosaurs have momentum. Thompson explains “Many of the

animations used for attacks have two distinct sequences: there's the wind-up, which

telegraphs the attack, followed by the big finish where the damage is dealt. … the system

controls the current locomotion of the machine, blends movement or attack animations to suit

at specific points and then ensure that the machine lands or stops in the right at the finish”

(Thompson ‘part 2’). The system Thompson mentions is another level of AI – a special kind of

programming that makes transitions between animations seem natural and look smooth. The

individual robot dinosaur senses or is told of the existence of the player, they are enraged by

this as all enemies are, they decide an appropriate course of action, and a special animation

system makes their body look good while they’re doing it.

The damage system of HZD means that enemies can be affected to a greater degree than

enemies in similar games. Robot dinosaurs are comprised of components that the player can

destroy one by one. The player can attack one dinosaur, blow off its gun turrets, and then leave

it crippled, if they like. The components the player is able to destroy are always a kind of

weapon, armour, or storage, so damaging these bits never results in the dinosaur limping or

losing any quality of life. However, the order in which the player takes out components

changes the course of a fight as enemies lose attacking abilities.

Nonplayer characters in Horizon Zero Dawn do not fight each other, but in some open-world

games, like Skyrim or Ark: Survival Evolved, animals will attack other animals both with and

136

without intervention by the player. They are enemies to each other, as well as the player. This

is exciting to stumble across, and gives a sense of the enemies having lives and concerns

independent of the player. But it also confirms their essential nature as angry, antagonistic

beings who can only solve their problems through violence.

This is taken to the extreme in the fighting game M.U.G.E.N. (this acronym does not stand for

anything). M.U.G.E.N is a game engine for which players can create their own characters.

Players can then battle against characters created by other people. Players are responsible for

drawing, animating, and programming moves for their avatars. These avatars are often based

on other game characters, cartoon characters, actors, or pop cultural icons like Ronald

McDonald. The joy of M.U.G.E.N lies in seeing two figures from different media interact, and the

challenge lies in programming a character that can successfully win battles. In ‘Watch Mode’,

both avatars are controlled by the computer – they become nonplayer characters. The ‘player’

has no input and just loads up two characters and watches them have at it. Because the

computer does not have to keep pace with a human, these battles are often lightning-fast.

Inevitably, M.U.G.E.N became an arms race of sneaky programming (Cybershell). Players began

making characters that could not be hit, that attacked instantly with massive weapons, that

could overwrite files on the other player’s computer, or could install malware. Battles with

these overpowered characters would sometimes just last a frame, or show a screen instantly

full of glitchy explosions and lasers. Because both characters are NPCs, controlled by AI

programming, the battles of M.U.G.E.N are between the computer and itself. The battles of

M.U.G.E.N are an example of NPCs engaging in their own affect loop without direct human

involvement. Nash notes that “Such a process [of a computer affecting itself] requires a

conscious acknowledgement of the difference between data-as-data and data-as-display” (14)

and although ultimately “the semi-autonomous behaviour of the agents can never be said to

be completely independent of the user … it is always the user that is playing the work (even if

not totally in control of the work)” (20), such a system can be said to produce “autonomically

aleatoric” (20) affect in players and characters; affect that is not deliberately designed but

something more haphazard and wild, that foregrounds the gameness of games and the

characterisation of characters.

137

HUMAN ACTION

It’s not too controversial to say that there are lots of kinds of computer games and players can

do all sorts of stuff in them. The mechanisms of how players fight and damage enemies are

varied. But there are commonalities: players and nonplayer characters are in opposition,

players and nonplayer characters cause each other pain (in a Spinozan sense), players and

nonplayer characters have different fighting actions, and the fights they engage in have a

beginning, middle and end.

In the open-world RPG The Elder Scrolls V: Skyrim, the player can use a number of fighting

tactics, depending on their location, their level, and their mood. Before all of that, they have to

perceive or be perceived. If a player detects an enemy before they themselves are detected,

they have a number of options. They can attack the enemy from a distance using long-range

weapons, keeping their own character out of harm’s way. They can sneak up close to the

enemy and deal damage before the enemy knows what’s happening. They can walk straight up

the enemy and punch them right in the nose. They can run in the other direction and avoid the

enemy entirely. The circumstances of perception involves a judgement on how the enemy

should be acted upon.

In Skyrim, the player has three kinds of weapons at their disposal: physical weapons like

swords and bows, magical abilities like fireballs and frost runes, and yelling, a sort of magical

projectile. These weapons are further divisible into categories like short-range and long-range,

slightly damaging and massively damaging, or cool and boring. But what the player will do

with any sort of weapon is hit the enemy’s body. In Skyrim, the player cannot ruin an enemy’s

reputation, steal their pants, or ask them if they’ve finished their thesis yet in order to crush

them mercilessly. They must touch their body. This refers back to the earliest games, where

the narrative premises were sporting (Pong, Soccer) or military (Space Invaders, Spacewar!,

Computer Space) contests of the body involving feats of physical strength and eventual

overpowering of one over the other.

138

The reception of blows may be presented to the player in a number of ways. In the Sonic

series, a single blow occasions death, but in other games enemies usually hang on a bit longer.

Because it is important to the player that they hit the enemy, acknowledgement of the blow is

welcome. These acknowledgements can come in the form of sounds, particle effects,

animations of the player-character or enemy, a reduction or increase in an enemy’s health-

related status bar or number, pop up text, counter-attacks, haptic feedback in a game

controller, blood splatter on the camera, or all of the above. In Street Fighter II, hitting an

enemy produces a punching sound, a recoil or prostrating or blocking animation of the enemy

character, a flashing animation around the area that was hit, and a reduction of their health

bar. If the player character has hit the enemy with a projectile, that projectile bounces off or

dissipates.

Similarly, winning or losing must be definitive in order to produce unambivalent affect.

Performance has a double meaning in games, and this double meaning holds the key to the

player’s place. A player performs in two ways. They act in a role like an actor plays in a play (in

her thesis Greer has noted the strong similarities between computer games and theatre). They

follow a script for their character (whether directed to or not (Dominguez et al.).) But players

also perform like a car performs, according to a certain expected standard of physical

functioning. This second meaning connects to mastery, though in a slightly difference sense to

how Plumwood uses it. The orientation of both performance and mastery matters. An actor’s

performance is created with an audience in mind (Greer ‘thesis’ 5), it is projected outward. A

car’s performance is something that can be measured, it is an internal quality of the car itself.

A person can be a master of slaves, their power directed towards others. Or they can master a

skill, without anyone else around. The definitions aren’t exclusive: an expert driver can make

an ordinary car perform well, an expert in a skill can leverage their expertise to boss others

about. Players of games must perform to a certain standard (to be able to make a jump or get

through a boss fight), temporarily inhabit a character role, and entertain someone (if only

themselves). They must overpower other characters, but must also become proficient in

certain actions.

RAIN WORLD

139

While most games with enemies involve some combat, others exist where running away and

hiding is the whole point. One such game is Rain World, a 2D third-person platforming

adventure about the post-apocalyptic adventures of a small white cat. Outside the god-like

terraforming / terrarium genre, few games attempt to create an ecology out of their

characters, in which all creatures have interdependent relationships. Although the creatures of

Rain World are sparse, they are each adapted to their environment and have a role to play in

the world and towards each other. And although their character designs and unique animation

style contribute to their believability, a large part of their verisimilitude comes from their

relationships towards each other. In Rain World, the player plays a character rather low down

on the food chain. The cat, although it can catch small creatures like bats and slugs, is itself

prey to lizards, vultures, scissor birds, giant centipedes, earwigs, spiders, leeches and

carnivorous plants. And these predators also fight and prey on each other – the player-

character is not at the permanent centre of their desires but rather a convenient snack in the

moment.

In the beginning of the game, the player can only hope to avoid or outrun their many enemies,

and does not know how to fight. The enemies are insatiable, quick, and often unforeseeable,

and the player-character gets killed in hundreds of gruesome ways. But both the player and

the enemies learn new techniques over time, and learn each other’s tricks. Based on their

actions in the game, the player earns a reputation among their enemies, who will confidently

attack, fearfully avoid, or nonchalantly ignore the player-character as a result. These last two

happen so gradually that the player has time to appreciate the slow turning of the tables. The

enemies of Rain World are ambiguous as they do not have any personal animosity for the slug-

cat, and are attacked by other predators as often as they attack the player-character. The

predators of Rain World are a frightening bunch, not least for the reason that they are in the

right.

The organising principle of Rain World is its rain. It operates as night-time does in other

games, being a dangerous and regular occurrence from which the player must seek shelter.

Every five minutes or so, bone-crushing rain thunders down, killing everything that isn’t

hidden in a bunker or pipe. The player must hide from the rain in special rooms, and

140

constantly keep an eye on the weather while they’re out. The player must also keep a watch

for enemies. The most effective way to survive enemies in Rain World is to run away from

them – the player can eventually learn to spear them but this takes hours of practice, and

some enemies are unkillable. Fleeing and avoiding enemies is a much safer proposition.

Rain World was inspired by the lead designer’s observation of rats in a railway tunnel

(Jakobsson and Therrien). Regular trains means rats have only a limited time to search for

food and socialise before needing to find shelter again. Luck is a key element. Enemies are just

as capable as player-characters at moving around the environment (and often better), and are

just as sneaky and daring. The player is often surprised by enemies popping up out of

nowhere and snatching them. While some have called this gameplay unfair or unbalanced (see

the comments in Jakobsson and Therrien above), the result of the enemies’ unpredictability

and skill is that the player becomes aware of their otherness from the enemy’s perspective. As

Plumwood put it while recounting her experience of being eaten by a crocodile: “A subjectively

centred framework capable of sustaining action and purpose must I think view the world

‘from the inside’, structured so as to sustain the concept of an invincible, or at least a

continuing self; we remake the world in that way as actionable, investing it with meaning,

reconceiving it as sane, survivable, amenable to hope and resolution. The lack of fit between

this subject-centred version in which one's own death is unimaginable, and an ‘out-side’

version of the world comes into play in extreme moments. … when my consciousness had to

know the bitter certainty of its end I glimpsed the world for the first time ‘from the outside’, as

no longer my world, as raw necessity, an unrecognisably bleak order which would go on

without me, indifferent to my will and struggle, to my life as to my death” (1995 30).

The killing of the player-character of Rain World is not unlike something from a nature

documentary – the cat is caught, shaken like a rat held by a terrier, and carried off. It is not a

sentimental death, nor a particularly momentous one. But for a few minutes after the player-

character has been killed, the camera follows the killer. The player can watch as the lizard or

bird or insect or plant drags their cute cat through tunnels, fights off other hungry creatures,

accidentally drops the cat, picks it up again, and otherwise treats the player-character like a

piece of meat. (This can actually be quite useful for discovering new areas or learning more

about the predator behaviour). The totality of the player-character’s objectification by the

141

other characters of the game is sobering, and reinforces the player-character’s place in the

world as one-amongst-many, a cog in the uncaring machine, a small part of a larger whole. The

lack of ceremony around the cat’s death is affecting not only because it interrupts the

gameplay, but because it shows how little regard others have for the plucky little hero, despite

it’s charming looks and courageous deeds, and highlights how little control the player-

character, and by extension the player, actually has in the world, by virtue of its ecological

niche and how it is perceived by others.

DYING

ENEMY DEATH

Enemies literally fight to the death, as their instantiated form is deleted when the player deals

the final blow. Like Star Trek personnel in a transporter, their body is destroyed and an exact

duplicate is made when it’s time for them to exist again. This operation is usually handled by

the enemy character’s code – upon a certain event happening (the player contacts the enemy;

the enemy’s health hits zero; a time has elapsed), the enemy destroys itself. As one

programmer proclaims in the Unity game engine forums, “Let the enemy decide if it's dead or

not” (juicyz).

Damage systems can be simple, where a single touch instantly occasions death, or a more

complicated one of blow velocity, weapon level, accrued hit points, health potion, and multiple

player lives. In Sonic games, Sonic is so powerful that one prick from his blue quills flattens

robots. In Horizon Zero Dawn, damage is a more complicated affair, but the player can monitor

the process of decay by counting the body parts they blow up.

When an enemy character dies in a modern game, they are more often than not deleted from

the scene entirely, and replaced with a new model of their own dead body. This happens so

142

quickly that the player cannot see one body disappear and another reappear, but a smooth

transition between the two can be difficult to animate, so sometimes there are explosions,

bloody spurting, or some other dramatics to hide the changeover. The player sees the enemy

as changing from one state to another, rather than being replaced. Replacing the dead enemy

with a replica of its own corpse means the game program has to work less hard: the enemy

corpse can be a simpler model to render and it doesn’t need to have any of its complicated

behaviour scripts. A dead character can also behave differently than an alive character, like

breaking into parts, being filled with valuable loot, or being a floppy physics ragdoll. Rather

than making an enemy that looks and behaves differently in different circumstances, it is much

easier just to make two different versions. The Unity game engine manual describes how a

game designer should program “a vehicle, building or character, for example a robot, breaking

apart into many pieces. In this scenario, the example script deletes and replaces the complete,

operational robot Prefab with a wrecked robot Prefab. This wrecked Prefab consists of

separate broken parts of the robot, each set up with Rigidbodies and Particle Systems of their

own” (Unity Technologies). Continuity endures in the mind of the player, but to the game (and

the NPC), the alive character and the dead character are completely separate things.

In older games, like Final Fantasy VI, dying enemies fade from view for the same reasons. In a

fantasy game this can easily be justified in the narrative, as supernatural creatures would

naturally dissolve into into ether, and a high-powered magic attack would vaporise enemies

into dust, etc. But a disappearing enemy body also signifies that the enemy character is

insignificant to the player once the fight is over, that it’s only a character through its actions.

Once an enemy cannot act, it may as well vanish.

Death comes to all NPCs, either directly at the hands of the player, when the camera is no

longer looking at them, or when the game is exited. They don’t seem to mind. Their

mindlessness is justification for their domination, but it also puts them in interesting contrast

to Spinoza’s conception of human desire. Death can only come about from an external cause,

Spinoza asserts (136), and things don’t want to die (137). Taking Armstrong’s interpretation

of ‘external’ in the sense of foreign or antithetical rather than physically outside, this accords

with Spinoza’s further claim that everything endeavours to persist in its own being (137). This

starts in a kind of Newtonian laws-of-motion discussion of inertness, but soon turns into a

143

principle of human psychology specifically. As Manning points out, Spinoza’s writing on

indivisible universal substance and his writing on essences is difficult to reconcile (7), so a

Spinozan position on nonplayer character death is hard to articulate. The endeavour to persist

is the basis for all desire (200), desire is the basis for pleasure and pain (202), and is also the

essence of the individual (173), and the rational pursuit of pleasure and negation of pain is the

basis of virtue (204), so the continuation of existence is inextricably tied to individuality (as

Spinoza says, to want anything at all you first have to want to exist in the first place (203)).

This leaves nonplayer characters, and their deaths, in a tricky position. They have no

conscious striving, as humans do, so the continuation of their existence is down to Spinoza’s

first identified motivation, inertia. They keep existing until something makes them stop. They

are more akin to a stone than a human, and what Spinoza says about stones is that they, like all

of us, “are all determined by external causes to exist and operate in a given determinate

manner” (390). This seems to be a good analogy to nonplayer characters so far. Spinoza goes

on to say that a stone, if it suddenly developed consciousness but no power to act, “would

believe itself to be completely free, and would think that it continued in motion solely because

of its own wish” (390). A conscious stone would feel like it was acting freely, even if it was

compelled. Imagining a stone as a kind of small human, though, fails to recognise its otherness,

as Plumwood would say. To understand nonplayer characters’ mindless relation to death

requires another perspective.

Continuing the analogy of nonplayer characters as a type of rock, Plumwood’s essay Journey to

the Heart of Stone investigates the materialism of stones as a site of re-enchantment. Much like

nonplayer characters, rocks have no feelings and destroying them can’t really be thought of as

killing them. For Plumwood, the analogy can only go so far though. Unlike rocks, nonplayer

characters are designed objects, so acknowledging them as others is a different project than

the liberation of rocks from western rationality. Neither nonplayer characters and rocks are

human, (or, in other words, both are non-human), but because they are designed, nonplayer

characters are caught up in different webs of meaning, and cannot be identified with the

“more-than-human world” (19) of nature. Plumwood writes “Stones are at the bottom of the

pile in moral consideration according to most versions of environmental ethics, not even

registering zero on the scale of ethical attention – unless they are lucky enough to get into the

category of cultural objects by being inscribed by humans or somehow attracting their

144

cultural attention, when they might be held to have ‘aesthetic value’ or ‘heritage value’” (21).

Nonplayer characters have achieved this level of consideration by virtue of their being human-

made.

In a Plumwoodian sense, because enemies have a kind of death, they have a kind of life to

them. Nonplayer characters are beings of dependencies. It takes a lot of work for them to

begin existing, and they need a lot of things around them to work in order to keep existing.

They aren’t continuous, they are changeable, and they are interconnected. Western rationality,

Plumwood argues, sees death as a defeat and a severing of continuity (Plumwood 1993 99).

The otherworldly soul and its place in an eternal perfect realm is the purported solution to

this problem. In contrast to western thought about death (which she argues that Plato’s

writing in particular exemplifies), Plumwood sees death as a kind of recycling (Plumwood

2007 32) of materials from one form into another, what Nash would call modulation. In this

sense nonplayer characters are bad Platonic subjects and good Plumwoodian subjects. Their

purpose and meaning is in their action, and their existence from beginning to end is a

coherence of multiple forces. Although they do not have bodies, they are changeable things,

and so are more like rocks than spirits.

PLAYER-CHARACTER DEATH

The death of the player-character has a significant effect on the enemy nonplayer character, as

this event usually resets the game or the game level. A game enemy’s successful attack is also

its suicide. It does not feel joy in victory, though the event of the player-character’s death can

be affecting and can instigate new behaviours (a sort of Spinozan pleasure). A nonplayer

character that has killed the player character might gloat, loot the corpse, or return to an idle

animation, before the game or level resets. In other cases, the enemy might remain oblivious,

attacking the player-character’s corpse or the empty air. The player-character itself might

perform an animation, as Meretzky describes in Crash Bandicoot: “Many of [Crash

Bandicoot’s] death-moves were extremely well done and cute, such as his death by fire, where

he becomes a pair of eyes in an outline of charred bandicoot, which almost immediately

145

collapses into a small pile of ash where the two eyes fall onto the pile of ash, blinking in

surprise. It's an animation worthy of Wile E. Coyote.” (Meretzky 2001)

Permadeath is a game mechanic whereby a game ends if the player-character dies. If the

player ever feels like playing that game again, they have to start over from the beginning.

Roguelikes and roguelites, games where the player explores randomised dungeons, tend to

use this, and as a consequence enemies in those games tend to be simple, so the player does

not have to waste a lot of time learning about them. In Downwell, the player-character falls

down a well, trying to avoid increasingly abundant enemies. If the player-character touches an

enemy they die, and are returned to the top of the well. The challenge of Downwell is in dealing

with enemies quickly, either by navigating past them or killing them. Death cannot be blamed

on the enemies, as they’re basically just animated obstacles, but rather the player must blame

themselves for not performing well enough. The enemy characters are so basic that the player

cannot say they were outclassed, only that they themselves slipped up.

In other games though, death is not the end. Being able to die and come back to life means that

players can learn about enemies through trial and error. In Hollow Knight, the death of the

player-character results in a new nonplayer character – the player-character’s own angry

ghost. This ghost floats around in the place where the player-character died, and holds all the

money the player had collected up until their character’s death. The player must fight their

ghost and win in order to get their money back. The player doesn’t become their own worst

enemy exactly – the ghost puts up a fight but isn’t that powerful – but the enemy only exists

because the player-character died. If the player has a perfect run-through of the game, they

never encounter this enemy.

This feeling of completion is subverted in Hades, a game about the Sisyphean task of trying to

escape the underworld of Greek mythology. When the player-character dies in Hades, they

always return back to the beginning of the game, and must fight their way back up through the

levels of hell every time. Every time they do so, though, the characters they encounter

acknowledge the player’s repeated failures and remark (sometimes snidely) on their progress.

Progress can only be made, in fact, by dying and meeting characters again and again. The first

boss the player encounters is one of the three Furies: either Megaera, Alecto or Tisiphone.

146

After the player dies a few times, they fight one of the other Furies, then two Furies together.

The dialogue the player is presented with, the moves the Furies make, and the difficulty of the

battle, is dependent on how the player has performed in their previous lives. Rather than

starting over from a clean slate, the bosses remember their previous battle, and this affects

each subsequent fight. All games involving player-character death are repetitive tragedies, but

Hades makes it overtly part of the interpersonal dynamic.

Dynamic difficulty adjustment also comes up in more contexts than racing games. Most games

in which the player-character can advance in levels or ability uses some difficulty adjustment

to increase enemies’ and friends’ abilities at the same pace, though “some forms of dynamic

difficulty adjustment are less about controlling the overall difficulty of the game, and more

about making sure that the player does not experience long strings of luck (good or bad) from

the game’s random number generators” (Compton). In this case, how players and NPCs can

affect each other depends on how they have affected each other so far. In one game, Compton

observes, “If the player has lost teammates, or has missed several times in a row, their odds of

hitting [an enemy] will go up”. In this way, the abilities of NPCs are always tied to the abilities

of the player-character.

FEELING

NONHUMAN FEELINGS

The enemy’s body, a coalescence of code, player and visual presentation, locates its sense

apparatus in its code, and senses other code. It does not sense its environment in the same

way that a player senses their environment, but it is programmed to emulate a representation.

The player might have the impression that the enemy reacts to the sound of their player-

character’s footfall, but the enemy either receives or seeks out at superhuman speeds the code

that precipitates the sound. Some animals, like the platypus, also sense the electrical signals of

other beings. They rely not so much on sight and sound when hunting, but the detection of

147

signals that are imperceptible to humans. The platypus, as Eco points out, is an

incomprehensible thing to someone who has never seen anything like it, and so is its way of

viewing the world to someone who has never experienced anything like it. Like nonplayer

characters, platypuses are often described as chimeras and unsophisticated. But this has more

to do with the boxes humans want to fit them into than the creatures themselves. Trying to

understand the platypus or the nonplayer character can quickly turn into a discussion of

language and categories and the limits of human systems of logic. Platypuses are elusive and

slippery even when they’re hypothetical. Eco’s use of the platypus is a springboard to talk

about Kant’s ideas of perception and identification; their exceptional otherness makes them a

convenient example. To consistently de-centre the human and centre otherness is a difficult

task. Nonplayer characters’ exceptional, designed otherness makes them good case studies,

because they are both human-made and nonhuman.

Game enemies do not avoid the player, struggle to escape them, or scream when they see

them; that is, they do not enact fear. They instead only act as if they were angered, curious, or

compelled. They are drawn towards the player rather than away from them; they are engaged

rather than avoidant. Even when injured, an enemy generally keeps attacking, or recoups for

only a minute, rather than admit defeat. But even if enemy characters do not perfectly enact a

full range of familiar human emotions, that does not mean that they do not have feelings of a

strange, digital sort, if we take emotions to be awareness of affect (Spinoza 130).

Enemies are affected by humans not only in ways that involve modifications of their body, but

in a way that involves the modification of the information that they use to act, or, their

knowledge of the world. They are affected first by the player starting up the game, or the

player character’s nearby presence spawning them into the game world. What was dormant

becomes an active subroutine, a small computer program that can measure, store, and modify

variables. It can sense the world and itself by asking for information from the game world, and

it is fed information from other subroutines. It is then constantly aware of the player’s

presence, and behaves accordingly, whether lying in wait for the player to come across it, or by

moving towards the player. Once within a certain distance from the player, it attacks, changing

its animation and registering the amount of damage it and maybe the player has received. It

might sense and record the blows the player has dealt it. Occasionally these numbers trigger

148

animations or behaviours representative of injury, but usually the enemy fights perfectly

furiously until destruction.

The enemy character, insofar as it perceives anything, perceives it through changes in its code,

recalling Spinoza’s assertion that nothing external can be perceived except through

modifications of the body (105). These perceptions trigger the enemy to act in different ways.

A boss in Final Fantasy VI does not know the player exists until the player actually touches its

hit box, and sits quietly waiting to be activated. A player can dance all around the boss, but

until the boss is touched it is ignorant, unaffected and inactive. Since the boss cannot perceive

the world except through its sense of touch, it cannot be affected by the player any other way.

Once the boss registers the player’s existence, however, it springs into action.

[Image removed due to copyright restrictions]

Fig. 19: The player near Final Fantasy V’s WingRaptor, who cannot yet perceive the player.

(author screenshot)

[Image removed due to copyright restrictions]

Fig. 20: Once the player-character touches the WingRaptor, a dialogue box pops up, and the

player is whisked into the battle screen. (author screenshot)

[Image removed due to copyright restrictions]

Fig. 21: The battle screen for the player’s fight with WingRaptor. (author screenshot)

Spinoza’s work on emotions, written to describe affect in humans, break down somewhat

when applied to the inner lives of computer game characters. An enemy does not endeavour to

persist in its own being (Spinoza 136). Enemies can become stronger or weaker, but this does

not result in pleasure or pain as humans know it. Even though enemies register changes to

their state, and this information can change their subsequent behaviour, emotions are not

something that can be ascribed to them. If “emotional response is subsequently seen as a

149

retroactive narration of affective response”, as Nash puts it (19), nonplayer characters do not

have the ability to tell that story to themselves. But certain affections do result in the enemy

being more or less active. So instead of finding analogies for human emotions in computer

code, it makes sense to say that recording, rather than emotion, is an enemy’s awareness of

affect. Nash’s concept of modulation applies here, as it describes the process by which digital

data moves and changes:

Modulation is used here to mean the process of changing some phenomenon from one

register into another, for the purpose of storage, transmission and display. The term

also resonates with Deleuze’s (1992: 3) sense of modulation as ‘like a self-deforming

cast that will continuously change from one moment to the other, or like a sieve whose

mesh will transmute from point to point,’ as well as the musical sense of changing key,

the electronic sense of changing a signal with another signal or the social sense of

changing one’s tone of speech according to listener or circumstances. All these share

the characteristics of both intentionality and change [emphasis in original] (2).

So affect, instead of resulting in enemies having feelings, results in an intentional change, or

modulation.

Spinoza takes care to inform his readers that his propositions only apply to humans

(particularly, rational adult men) (171), but seeing the ways in which his definition of

emotions break down when applied to computer game characters helps to illuminate certain

aspects of their inner worlds. Computer game characters are beings who are designed to

mimic human emotions (and the emotions of elves, aliens, fluffy cats etc), but do not internally

operate like humans. In Spinoza's terminology, their body does not have an idea of a mind. It

seems absurd to talk about computer game characters feeling real emotions when they can’t

even fake them very well. But characters are modulated by the player and by their

environment into new forms. Their ability to act can move up or down on a scale. They can be

depressed – that is, have a decreased power to act (or more accurately, their ability to act is

suppressed). They can be elated, or have an increased power to act. They can be uplifted or

reduced in standing. In this sense, game enemies can be said to feel.

150

LIM

The world of LIM is a world in which difference is punished. The protagonist of LIM is a little

square (not in the uncool way, but the four-sided way), which the player must navigate

through a simple labyrinth. The game’s enemies are squares in both senses. They are

monocoloured while the player-character square is constantly flickering many different

colours. They attack the player-character if the player-character is a different colour to them.

The player can stop flickering colours and blend in with the enemies for a time, but this comes

at a cost to speed and navigability – the player-character moves slower and the 3rd-person

camera zooms in to focus intensely on their body. The player must choose between curtailing

their own movement or having others do it for them. Though this simulated interaction of

squares and nonconformists, a story is told in mechanics.

[Image removed due to copyright restrictions]

Fig. 22: An arrangement of squares spelling “LIM” (author screenshot)

To even identify which of the squares is them, the player must break up the game’s title

(Figure 22). To become an individual in this game and know who you even are, the player

must remove themselves from others. LIM does not present this creation event as

unproblematic. By becoming an individual, the player undoes a meaningful structure and the

rest of the game is concerned with fitting in again.

[Image removed due to copyright restrictions]

Fig. 23: The player-character square (brown) moves into view of an enemy square (blue), and is

attacked. (author screenshot)

LIM is in top-down 3rd-person perspective, as if the player-character were under a

microscope. Enemies do not kill the player-character but knock them around, making the

game disorientating. If the player blends in, the camera closes in on the player-character until

151

it almost fills the screen, and the player hears banging and sees their character quivering, even

though enemy characters are not attacking it.

[Image removed due to copyright restrictions]

Fig. 24: The player-character fixes itself blue, and the camera zooms in on it. (author screenshot)

The player thus has two choices: to be different and expect harassment, or to blend in and

inhabit a claustrophobic, paranoid state. They cannot win, only endure. Even blending in with

the enemies fails to result in camaraderie, only indifference. The player’s choice is to be

attacked or ignored, and they enact this choice through self-surveillance. LIM keeps the

enemies’ perception of the player-character at the forefront of the player’s thoughts. Even

though the player has the power to change themselves, this power is used in reaction to

others, with the desires of the enemy in mind. Even the labyrinth walls are made of squares

the same size as the characters, resulting in uncertainty of whether they are architectural

elements or silent observers. This blurring of the non-player-character, or NPC, role, and that

of environmental objects is significant as it calls into question whether boundaries are really

mechanical or social in nature. The limits of playable space may be walls, or they may be other

people.

Anticipating and managing the feelings of enemies is central to LIM’s gameplay. In trying to fit

in with the enemies, players experience a kind of power that Foucault describes as “power …

which categorises the individual, marks him by his own individuality, attaches him to his own

identity, imposes a law of truth on him which he must recognise and which others have to

recognise in him. It is a form of power which makes individuals subjects” (781). LIM lets

players experience the imposition of subjectivity, and encounter the difficulty of trying to

refuse it. Desubjectification, or the project of refusing who we are (Foucault 785), involves

navigating Foucauldian and Spinozan conceptions of power. Players can disempower enemies

in the Spinozan sense that they can stop them attacking, but they cannot escape power

relations in the Foucauldian sense, because power shapes the whole game experience.

152

Whether players of LIM feel like they are gracefully adapting to an oblivious population or

fearfully submitting to a hostile crowd, the perception of the enemies is key to their activity.

Players must work to keep enemies placid, but in doing so they elevate enemies’ feelings

above their own. The centrality of enemy feelings puts players in a subordinate position,

where everything they do is in relation to them. The point of the game is to not affect the

enemies, and in doing so they become the centre of the player’s world.

HUMAN FEELINGS

When research is done on computer games and feelings, it is usually done on whether combat

in computer games makes people more aggressive and prone to violence in everyday life

(Anderson; Anderson and Bushman; Dill and Dill; Ferguson; Gentile et al.; Sherry). Suffice to

say, results are mixed, but overall, they probably don’t for most people (Ferguson). This topic

has been covered thoroughly in other theses and papers, so this section will focus on other

feelings nonplayer characters might evoke in human players.

Although enemies are ultimately designed to be defeated by the player, the enemy still has a

good go. They act without hesitation, a programmed certainty in their movements. In being an

obstacle to the player's goals, they are often violent, persistent, and ferocious. They demand to

be attended to in precedence to the player's larger aims. They not only attack the player but

catch their eye and hold their attention. In games where players have goals to achieve, quests

to complete, things to do and places to go, enemies are a distraction, an encumbrance. But

fighting and overcoming these obstacles is often more enjoyable to the player than achieving

the end goal itself. Indeed, the bulk of a player’s time playing a game might be taken up with

conflict. Enemies seem to desire. The more single-minded they are, the more compelling

players find them (Signoretti et al). The player is the centre of the enemy’s world, and they

want to be the centre of the player’s world. And so the main action that enemies take towards

humans is attention-seeking, and genre and narrative merely inform the methods the enemies

will use (see Bright’s catalogue of all actions enemy characters can take in ‘retro’ games).

153

Despite advances in enemies’ more complex and smooth movement (Boulic et al), ability to

grow and adapt to player action (MartíAnez-Arellano et al), more human-like behaviour (Yuka

et al), a larger quantity of animations, and greater perception of events and surroundings

(Arrabales et al; Signoretti et al), interaction with enemies remains relatively simple. The

player must touch, sometimes with a tool, sometimes in a specific place, sometimes with a

certain frequency, the enemy’s body. And the player too, finds that the enemy desperately

wants to touch their character’s body. If this is starting to feel a bit sexy, you’re not wrong.

Enemies and players are passionately compelled to act upon each other to reach a dramatic

climax. (As Anderlini-D'Onofrio has noted, the dramatic structure of climax itself is informed

by Aristotle’s ideology of hierarchy (31)). The point of an enemy encounter is the creation and

relief of frustration. Enemies pop into existence temporarily for the thrill of making them

nonexistent again. Fights with enemies follow a familiar structure of initiation, rising action,

and climax (and falling action and catastrophe if the player loses).

Predictability is important for a fight to be satisfying. As this review for River City Girls

complains, “Even enemies of the same exact type do not stay down a consistent amount of

time. Also, when you're trying to stomp, it may or may not actually do damage. Sometimes

you'll stomp and just nothing happens at all” (Dredd). What some players may feel when they

encounter nonplayer characters in games is disappointment. Game designers have various

desired emotional states they would like the player to experience (Schell 9), certain intentions

for the player’s mood. Good game design is the design of an enjoyable emotional rollercoaster.

As game reviews remind us, a game’s quality (and the progress of the games industry as a

whole) can be measured by how much they make players (deliberately) cry (GamesRadar

Staff; Giuliani and Baabuska; PCGamer). Anable observes that “The ability to evoke strong

empathetic responses, like crying, continues to be posited as the litmus test for taking video

games seriously.” (viii). What game designers want players to feel when encountering enemies

is not boredom, not sadness, not exasperation, but cool feelings like fear, anger, anxiety, or

umbrage. But as Isbister notes, these are more often feelings that result from circumstances

rather than the character themselves: “because there is no personal investment on the part of

the player in individual enemies and relationships with them, emotional moments arise from

brief encounters around combat situations—the first sighting of a new class of enemy, a

154

surprise ambush, the moment of success in killing an enemy, or the sting of being defeated”

(242).

Fogel et al. summarise the frustrations of designing ultimately disappointing game characters

when they state:

“The failure of AI methods to really create novel high-quality behaviours is legendary in

the entertainment software business. Programmers cannot conceive of everything that

might happen in a game, and when unanticipated situations arise, programmed

characters often act in ways that no human opponent would consider rational. The

main way that software developers have achieved any measure of success in

overcoming this is by iterative play testing, finding these weaknesses, and writing more

rules to cover them. Ultimately, this patchwork method fails, and games are released

even though the developer knows that the game characters will act contrived and will

not maintain the player's interest for very long” (1420).

But affect that designers do not intend can also be interesting experiences. Johnson enthuses

that “the range of experiences that can be described as emotionally or affectively frustrating

while operating game media are … quite diverse” (594), while Anable argues that games are “a

medium that seems to traffic more interestingly in the minor affects” (viii). That games are

sometimes more maddening than saddening, or annoyful than joyful, is an interesting

phenomenon. Spinoza reminds us that what affects one person one way will affect another

person another way, and the same thing will affect the same person differently at different

times (163; 208). When a player feels frustrated at The Last Guardian’s idiosyncratic camera

system, they might go about the rest of their day irritable and tense, or they might reflect on

how the wonky controls enhance the game’s themes of precarity. When a player finds

themselves repetitively picking and bottling peaches for hours in Stardew Valley, they might

find it a soothing meditation that relaxes them after work or they might find it an unpleasant

obligation that weighs on their mind. How games affect players at their most dull and ordinary

moments is just as important as how they affect players in their most intense and rare

moments.

155

More unpredictable opponents also makes for a more ambivalent gaming experience. As

MartíAnez-Arellano coyly put it, “human players tend to dislike very weak AI [enemy]

characters. However, when they are presented with characters that are roughly equivalent in

strength, their perception varies” (434). Enemies are not made to be liked but to present the

player with difficulties; while these difficulties have traditionally been those of timing and

dexterity, emotionally difficult characters can provide their own challenges. As players’

diverse reactions to Rain World and River City Girls shows, enemies that deviate from

expectation can make or break a gaming experience. If an enemy is unpredictable, boring, or

too hard, a player can come away from a game angry, upset at the game, or upset at

themselves. Fighting in a game is often an action-packed situation in which the player is

expected to move quickly and precisely. Failure to do this results in the player blaming the

game for not being fair, or blaming the self for not being good enough. In the case that the

player blames the game, according to Spinoza’s definition of the emotions (140) they are

expressing indignation at being failed by the game developers. In the case that the player

blames themselves, they are feeling shame.

Success in defeating a difficult enemy results in pride. However, if the player perceives that

they have succeeded not due to their own efforts, but to a rigged game design, they will once

again feel cheated. Managing the emotions of a player encountering an enemy, then, is a

difficult balance between making the challenge manageable, but making the player feel like

they managed it all on their own. In this dilemma, a player’s sense of progress is important. If

the player knows that they have just defeated an enemy that would have thoroughly trounced

them a mere 10 hours ago, they can feel that their success is the satisfying result of their own

hard work.

merritt k observes that player performance during a game can be used a measure of personal

worthiness, especially in family settings, and especially when their worthiness is already in

question. Post-gamergate, the games journalist Dean Takahashi uploaded footage of him

getting absolutely walloped by the game Cuphead, and re-ignited a debate about the point of

games journalism and the quality of games journalists (Tamburro). Common experience of a

game with peers can be a part of a game’s affective experience, and seeing Takahashi play

Cuphead badly brought up feelings for some about who counted as a part of a community.

156

Takahashi’s worth as a journalist was tied to his performance in the game, with some

questioning whether people should be allowed to become games journalists without being

very good at games (Tamburro).

Much has been written about the identification of the player with the player character, and the

emotions that go into such a bond (Lankoski et al; Larsen; Klevjer; Rehak; Sageng; SchroH ter

and Thon; von der PuH tten et al.; Yoon and Vargas). A Spinozan perspective would see the

player-character as a love object. A player-character is loved because it enables the player to

act within the game world, and because the player invests so much in their maintenance and

upkeep. Enemies are a threat to this love object. It is difficult to feel happy about a wolf

gnawing on a player-character on which you just spent an hour perfecting the hair colour. If a

player imagines themselves to be hated by enemy nonplayer characters (and since enemies

attack ruthlessly, instantly and specifically at the player, they certainly seem motivated by

hatred), they will quickly come to hate them back, because the thought of being hated is a

painful one (Spinoza 157). If the player believes that the hatred of the enemy is not justified

they will feel shame, but as Spinoza notes, that rarely happens.

But players do not have to initially hate enemies in order to perform violence towards them.

“This reciprocation of hatred may also arise from the hatred, which follows an endeavour to

injure the object of our hate. He therefore who conceives that he is hated by another will

conceive his enemy as the cause of some evil or pain; thus he will be affected with pain or fear

[fear being “an inconstant pain arising from the idea of something past or future, whereof we

to a certain extent doubt the issue” (176)], accompanied by the idea of his enemy as cause; in

other words, he will be affected with hatred towards his enemy” (157). By this Spinoza refers

to his earlier assertion that “we in no case desire a thing because we deem it good, but,

contrariwise, we deem a thing good because we desire it: consequently we deem evil that

which we shrink from” (156) and “in no case do we strive for, wish for, long for or desire

anything, because we deem it to be good, but on the other hand we deem a thing good,

because we strive for it, wish for it, long for it, or desire it.” (137). In this case, the hatred can

come after the player has done injury to the enemy, and comes because the player has done

injury to the enemy.

157

Hurt, it must be said, feels different when affecting a player-character body rather than a

human body. The hurt of a player-character body is registered in numbers or in bar form

rather than in the sensation of pain. The pain the player feels is their reduction in capacity –

they have to start the level again, they’ve lost time or gold rings, they’ve failed themselves,

they’ve made themselves look silly. This affect does not even have to be consciously recognised

– the beats in a game’s music can change a player’s heart rate (Cole). The player may find their

breathing changed, their muscles tensed, their reactions fast and twitchy, their thoughts

focused, their skin breaking out in sweat. Alternately, they may feel slow, easily distracted and

hopeless, if they perceive the challenge to be greater than their ability to overcome it. This

stress does not have to be processed as emotion but can remain bodily sensation.

SHADOW OF THE COLOSSUS

The underlying discomfort in attacking creatures who are just peacefully hanging out in their

own home is turned all the way up in Shadow of the Colossus (SOTC). The gameplay of SOTC

consists of fighting sixteen huge monsters, each more massive and challenging than the last.

Even though the monsters are hostile, aren't particularly cute, don't resemble living creatures,

and their deaths are the entire point of the game, fighting them is a morally conflicted, even

sad occasion. The colossi behave like other computer game bosses, but everything around

them – the music, the player rewards, the narrative – indicates that Shadow of the Colossus is

not a story about good guys and bad guys, but more about unfortunate guys and even more

unfortunate guys.

The player's purpose is to kill these sixteen monsters in order to magically bring a girl back

from the dead: a very noble purpose. But these creatures don't really want to be sacrificed,

and aren't wild about being killed. It is also clear that these are ancient, unique, and magical

creatures, and the player's motives are selfish. The calculation of sixteen deaths to reverse one

death comes to seem perverse. Huge monsters are traditional game enemies, but in the case of

SOTC, it is the player who is the enemy to the creatures, not the other way around.

158

Tom Cole’s investigation into the games of Team Ico (the game development company behind

Ico, Shadow of the Colossus, and The Last Guardian), shows that game designers can use very

simple mechanics to great effect (and great affect). He observes that at the core of Team Ico

games is a concern for others, and that their games use a number of specific techniques to

encourage players to care for characters, objects and events. Cole argues that concern for

others is “a necessary pre-condition of emotional response” (3) and the willingness of the

player to care for characters in a game creates the space for emotionally complex interactions.

Cole argues that Team Ico games are essentially other-centred, making the player care more

for nonplayer characters than their own player-character. Players can be affected by other

characters, he argues, when the game prioritises “fortunes-of-others (focusing on events

happening to others), attraction (focusing on an object) and attribution (focusing on other

agents) emotions” (9). SOTC, Cole argues, encourages the player to care about the colossi by

expressing these priorities in six ways: music, betrayal, individuality, time, exclusivity and

vagueness. Firstly, “music and sound are key to the establishment of ‘moods’ … which

predispose the player towards experiencing certain emotions linked to that mood, and to seek

out audio-visual cues to perpetuate that mood” (4). The soundtrack of SOTC is sparse, only

playing at intense moments such as the approach of an enemy or a threat to survival. The

music that plays is not the typical rousing chords of ordinary game battles; nor does a

victorious orchestral score accompany the slaying of a colossus. Instead, a slow, sad requiem

accompanies the battle, and a haunting, melancholic melody plays after the death of the

beasts. This plays into what Cole calls the “betrayal” of Team Ico games – actions that would

result in victory and elation in other games are presented as tragic, unjustifiable events. This

contradiction of success and mournfulness is confusing, and brings mixed emotions to an

event that in any other game would be five seconds of happiness.

Sad music when a colossus falls “is a rare example of a player being ‘punished’ emotionally for

achieving an objective, and amounts to a betrayal of the player’s expectation of ‘hard fun’. The

player feels the conflict between the ludic rewards and the emotional payback, and it is likely

to elicit an emotion rarely felt in videogames – guilt” (9). Although the player has done the

right thing by the logic of games – killed the enemy, completed the quest and saved the girl –

they are brought to account by the mood of the game, and made to critically reflect upon the

nature of their actions. Cole observes that “events that run counter to the audience’s beliefs

159

and desires can elicit just as much, if not more affect, than those that run parallel” (9). That

SOTC emotionally ‘punishes’ players for following standard game narratives results in a more

intense, and more complex affective reaction than if the game had followed convention. And

yet, this unconventional gameplay relies upon convention in order to work. Subverting

expectations would not be so powerful if these expectations had not been built up over

decades of game design. Lankoski and Helio’s ‘expectation of gaming’ – that the player will

want to see action – creates the untenable situation that the player must keep killing

magnificent creatures in order to advance SOTC. But the game makes it clear that this act is

deeply wrong. This ludonarrotive dissonance, or betrayal as Cole terms it, adds a very complex

layer to the game without demanding much more in resources to create.

BOSSES

So far we have considered players as actors and nonplayer character as reactors. But game

bosses, though they lie in wait for the player, though they exist only as an extra challenge, are

by design almost equal to them in skill. Game bosses have the potential to dominate the player

as the player has dominated others.

Bosses are special enemies. If they win, it is because they are better than the player, not

because the player has slipped up and failed themselves. They are more difficult to defeat, they

cannot be avoided, and they have unique character designs. Battles with them often occur in a

closed arena which the player cannot escape. Usually, the player must lose to a boss many

times before they learn how to defeat them. Compared to other enemies in the game, bosses

have a larger range of moves they can draw upon, and they can surprise the player by

switching between strategies. Bosses can also exhibit an extinction burst, where a new and

more powerful attack is unleashed after the boss has accrued a certain amount of damage.

Bosses, according to their name, have a big job to do. They usually have more narrative

justification than other enemies, and further the game’s story or contribute to its

worldbuilding. They are also gatekeepers, only letting the player access new information or a

160

new place after they have been defeated. A boss gives the player a more difficult fight, but the

rewards are also greater. They epitomise the idea that if the player strives harder, they will

gain more; that the grind will pay off.

At the end of Final Fantasy VI (FFVI), the player fights Kefka, a figure who has been taunting

the player throughout the game. He is a kind of unhinged, clownish character who attains the

form of a god by the end of the game. FFVI has a menu-based combat system, and all the

player’s and enemies’ actions have names. These names are usually pragmatic and descriptive

like ‘fire 2’ or ‘cure’, but Kefka’s actions are poetic: ‘calmness’ for instantly obliterating a

player-character, ‘goner’ for inflicting massive damage, and ‘fallen one’ which reduces the hit

points for the player’s whole party to 1. Kefka’s mission is to destroy the world bit even if the

player defeats Kefka, the world is still destroyed, and slowly falls apart. The game ends on the

sobering note that although everything is dying, the player at least got revenge. Kefka aside,

one of the pleasures of encountering a boss is a sense of finality. Bosses live at the end of

something, and so encountering a boss means the player has finally arrived, that their mission

is complete.

CONCLUSION

As River City Girls’ Kyoko muses:

[Image removed due to copyright restrictions]

Fig. 25: A player-character wonders why everyone wants to fight her all the time. (author

screenshot)

With all that said, players tend to be the more active partner in a game (Flanagan 280). Players

run the show and enemies react to what humans do. Humans, then, have a particular part to

play in initiating, maintaining, and ending, affective interactions. These parts have been

scripted for them by designers and afforded by mechanics, so the actions humans take aren’t

161

entirely up to them. Players’ actions are constrained by game mechanics like enemies’ are, but

because players have a wider range of actions available to them and a larger number of

decisions to make, human actions are significant in the development of affective relationships.

Enemies are common but shallow game characters. Although they are dissimilar to anything

in nature, the way the player behaves towards them is an extension of the logics that inform

familiar power dynamics. They do not mimic reality in the way they behave, but in the way

they are behaved towards. Where there have been controversies about enemy characters as

representations of particular types of humans, this is because they have been placed into the

same position as all other game enemies: peripheral to the player, who is the centre of the

game. And game enemies, like other beings subservient to a dominant other, are characterised

(literally, in their case) as a mindless, unintelligent, violent mass.

Players not only have the power to affect enemies in and out of existence, but are the central

focus of enemies' attention during their short life. They are easily distinguishable by

appearance so the player can quickly identify their particular type and react to them

accordingly. Unless they are a boss, they are not treated as individuals but as instances of a

class, all of which have the same universal attributes and motivations. Within a class,

individuals can be rarely distinguished, and so the player develops general strategies to deal

with these generic beings. The player knows what to expect from each enemy based on their

appearance, and knows that each enemy of the same type will behave the same. New,

unfamiliar enemy types, variations on types, large numbers of enemies, and tricky locations,

then, keep the player entertained after existing enemies have become familiar. Enemies are

not expected to develop as individuals, but to be replaced by something more interesting.

Enemies, then, are some of the most robotic, machine-like characters, but also some of the

most affecting.

Enemies are generic creatures who are tightly tied to specific places. Their design is repeated

and general but their body is singular and local. Although they are all clones, none of them

original, they become individual by belonging to a location, by starting their lives in a

particular place. They have spawn points which determine the precise location in which they

are to be born. They initiate at that place when the player is close by or their level loads. They

162

render into a visible form when a camera sweeps a game world into existence. Most enemies

do not venture too far from their spawn points. They love their homes and are creatures of

habit. In this way, these interchangeable instances acquire specificity, by existing in a

particular place and time, in the view of a human who assigns them meaning.

Enemies are exciting to interact with, and battling them is one of the most memorable and

rewarding parts of a game. Because the player may lose progress or assets as a result of a

battle, there are real stakes to interactions with enemies. A difficult or inconvenient enemy

may inspire hatred or fear in the player, but all enemies are designed to cause the player pain.

Enemies are stressful, and interacting with them requires concentration and fortitude. They

are characters who demand to be attended to, which is the extent of their selfhood.

Players affect characters through mechanics, but nonplayer characters affect players through

mechanics, narrative, association, sound, and self-reflection. While the appearance, behaviour

and role of the nonplayer character brings up all sorts of associations and connotations for the

player, nonplayer characters have no broader cultural context within which to judge the

player, and no ability to do so even if they did. But they are still affected through the player's

characteristics, and act as if they are aware of the player's place in the world. This is so the

player can feel more keenly their own place, but it also gives a sense that nonplayer characters

have a continuous experience and are somewhat involved in affective networks. But nonplayer

characters do not talk to each other, and the way they affect each other is limited. Instead, all

their knowledge about their world is programmed into them, or they recognise tags and

integer thresholds in the player's code.

Rendering visual bodies is expensive – it uses a lot of energy and draws on the computer's

resources more heavily. So in an open-world game where events can happen off-screen, the

game world still happens in a certain radius around the player, but nothing has a visible body.

It can be helpful to think of NPCs (and all game assets really) as all frolicking around naked

until the instant the camera is looking. When the camera looks at a new area, everyone hastily

renders their bodies on. Some game assets (usually landscape elements rather than NPCs) do

a half-arsed job and put on simple shapes, only changing into their fancy high-poly bodies

once the player gets within a certain distance. So in some games, nonplayer characters can act

163

without being seen by the player. They can affect and be affected, but their bodies are not

visible. The animations appropriate to the character's situation may still be triggered (in case

the player comes across the character while they're in the middle of doing something), but

without a body to be animated. In simpler games, only things that are shown on the screen are

active, and nonplayer characters do not exist as agents unless they also exist as objects of the

player's gaze. Players and enemies are always perceptible to each other. The enemy only

inhabits a visible body and has a full range of embodied behaviours when the camera is turned

to it. Such characters begin and end their existence dependent on the camera, which follows

the player.

164

CHAPTER FOUR: FRIENDS

INTRODUCTION

The game friend is an odd creature. It doesn't provide a challenge as enemies do, its purpose

(if not effect) is to make the game easier, and most of its functions could be replaced by other

mechanics. It seems counter-intuitive that a program designed to make life temporarily

difficult would include a subsystem for friendliness and helpfulness. However, games are not

entirely about killing zombies (J. Smith). And as Costikyan observes, interpersonal difficulties

can be just as challenging as fire-breathing dragons (14). If games are a designed emotional

experience (Jagoda and McDonald 176), then pleasant interactions can also be part of that

design.

The computer game friend is usually more of an employee than a buddy, and it has a difficult

job. They must stay close to the player, but not so close that they get in the way. They must

assist in battles, but not so much that they win on their own merits. They must be vulnerable

enough for the player to worry about their welfare, but only die at the most climactic moment.

They must also look different enough to the player character to avoid visual confusion, but still

be relatable enough to arouse the player’s sympathies. Ideally, a game friend is loyal but not

too dependent, funny but not too wordy, useful but not too capable, active but not an actor, and

universally attractive to any human who looks at them. They cannot embody everything about

friendship between humans in other contexts, because they are a computer game character in

a particular context. Some aspects must be left out for time, complexity, and genre. The kinds

of friendliness that are included in games, then, give an indication of what designers see as the

minimum requirements for friendship. The study of ally nonplayer characters in computer

games is also an examination of conventional understandings of friends.

Like other nonplayer characters, friendly NPCs are defined by their relation to the player.

Their appearance, behaviour, and story all flow from that role. For Schell, Isbister and

Warpefelt, game characters are always functional. They are made to fulfil a purpose, and so all

165

their characteristics point towards that purpose. Schell suggests that “a very useful technique

when coming up with the cast of characters in your game is to list all the functions that these

characters need to fulfil” (351). Isbister recommends that since “game character interactions

are usually short, targeted, functionally based” (225), designers should rely on conventional

social roles to quickly position and orient players to nonplayer characters (225). Warpefelt

(2016) further suggests that success for an NPC depends upon their verisimilitude in their

assigned role: “the NPC is actively involved in the portrayal of its role, and will act in ways that

are conducive to convincing the player that it is indeed in that role” (33). According to

Warpefelt, NPCs have particular tasks to fulfil (sell goods, be a challenging opponent, etc), but

unlike other objects they must perform that task with some theatricality (34); they should act

in a way that implies some limited, familiar personhood. Isbister notes that nonplayer

characters, no matter their role within their game, ultimately exist to serve the player.

“NPCs have a range of objectives in games, from unthinking service and loyalty to

the player (or his or her archenemy) to world domination at the player’s expense.

An NPC may have various abilities that are useful to the player. Beyond the ability

to provide physical or mental assistance in fighting, solving puzzles, or just in

learning the ropes of the game world, a friendly NPC may also be able to provide

moral support in achieving game goals: cheering, excitement, approval, and the

like. NPCs may also provide companionship for the player or may provide a social

motivation in the form of someone who needs rescuing. Even unfriendly NPCs have

abilities that improve the player’s experience, providing opposition and conflict—

both physical and emotional—that enhances the player’s experience. NPCs in

neutral roles can provide social validation for the player when they approve of her

or his actions and help spur better play when they boo a bad performance.” (228)

Although nonplayer characters all perform different actions, they all do it for the same reason:

to give the player a good time. To ask whether an NPC is a friend or an enemy is to ask

whether an NPC is a friend or an enemy to the player. This fundamental purpose of being,

though, is already at odds at what it means to be a friend.

166

Game friends, unlike game enemies, have roles based on relationships that occur between

beings in real life. Game friends are designed to be pets, mentors, dependants, best buddies,

employees and co-workers. Their conviviality lies in their familiarity. But these roles are

boiled down to their barest essentials – or more accurately, what designers believe to be

essentials, or just the qualities of amiability that are the most easily identifiable, animated and

programmed. Through this distilling process they develop their own spirits, preserving only a

tenuous connection to their real-life inspirations. These inspirations, however, still serve as a

common foundation for a player’s individual relationship (DíAaz and Tungtjitcharoen 7),

despite the game characters’ distant evolution from it. Friends cohere out of the pixels on the

screen, their performance, and ideas the player brings to the game (Simon 4). They cohere as

characters in the mind of the player, and as subroutines in the depths of the computer.

The unique but strangely familiar relationships players have with game friends start from role

recognition but develop through the actions players take. Friends do not touch each other as

much as enemies do, and this affects the immediacy of relationships. Dialogue and resource

provision are the main mechanics of game friendship, but are not particularly exciting in

themselves. Talking to a game character can be a challenge, but not a thrilling challenge like

punching them would be. Interactions with enemies are immediately rewarding because the

ways that players act towards them are enjoyable. But interactions with friends rely on the

promise of future rewards, like the acquisition of information or resources. Friends, then, have

to signal their friendliness in diverse ways rather than purely through mechanics. They cannot

rely on interactions with them being immediately affecting like enemies can. Yee calls such

interaction systems between Massively Multiplayer Online Role Playing Game (MMORPG)

players ‘social architecture’, but this term can also be used for interactions between players

and nonplayer characters. This architecture shapes how players relate to other characters –

the friendly relationships games can afford, and therefore social architecture can be said to be

part of a game’s mechanics.

Enemies are frustrating, terrifying or boring because they fight in frustrating, terrifying or

boring ways. Their characterisation is important, but even a grey cube (as LIM shows) can be

an affective enemy. Friends rely on familiar social roles, narrative, and character design, as

well as mechanics, to affect the player. The more enjoyable aspects of non-game friendships,

167

like rapport, mutual aid, belonging, understanding, trust, and hugs, are tricky to replicate

between players and characters in a game context, and the common tools players have to

develop this relationship are inadequate to the task. Players are affected by enemies because

of the ways they can affect each other, but players are supposed to feel friendly towards ally

characters despite the ways they can interact with them. However, this melange of role,

expectation, and mechanic creates a new kind of relationship that is not quite like friendship

between humans, but something else entirely.

Friends are usually more concerned with scripted story and plot than enemies are. Player

interactions with enemies can produce emergent narratives that are meaningful to the player,

but attempts to make emergent narrative with friends through AI stories have not been quite

as successful (Lenhardt). A game enemy may be able to be won over to the player’s side and

become a friend. A friend might betray, or be betrayed by the player. A friend can be affective

in their absence. Friends can be extensions of the player character, or they can be

incomprehensible aliens who just don’t attack the player. Friends can come and go, apparently

progressing in their own stories as the player does theirs (the Final Fantasy series is a good

example of this). This means that friends are concerned with time and progression more than

enemies are. Game enemies iterate over a game, levelling up in scariness and fighting ability,

but game friends aren’t replaced every five minutes with someone friendlier. They persist, and

share things with the player, even if it’s just space and time. Through these things, game

friends can come to mean more to the player than just a portable second inventory.

Many games have performed interesting experiments in the mechanics of friendship. In Ico,

the player-character leads a nonplayer character by the hand, relying on her to solve puzzles

and encouraging her when she encounters difficulties. In Spider, the player is subject to the

constant, intense gaze of a huge, benign monster. In The Last Guardian, the player cooperates

with their capricious pet chimera to navigate obstacles. In the Final Fantasy series, multiple

characters join and leave a group hiveminded by the player. These games can obviously blur

the boundaries between player and nonplayer characters. Sometimes, this blurring brings

more bodies into the being of the player character. Sometimes, it decentres the player

character and the player, making the maintenance of a relationship rather than the

maintenance of single character the focus of play.

168

THE DISAPPOINTING FRIENDS OF COMPUTER GAMES

There's a lot to be said about game characters who aren’t intensely affective. To have a series

of small forgettable interactions with mildly pleasant characters can contribute to a good

mood just as effectively as one spectacularly lovable character. The ability to interact with

characters without caring at all about their feelings can be a welcome relief from ordinary life,

especially an ordinary life that involves managing the emotions of others. Similarly, being the

centre of attention or occupying a respected, dominant position can be a pleasant and unusual

experience. Yet the relationships players have in games colours a whole game experience; as

Greer states, “the affective bond between the game player and the digital character in a game

is as vital and essential to the experience of playing the game as the emotional relationship

between the performer and the audience in a play or film” (Greer 2013 viii).

Some people find great pleasure in meeting a computer game character who doesn’t attack

them on sight. Some characters are so evocative of cherished concepts, so cute in their

designs, or are so helpful to the player in achieving their goals, that the player warms to their

presence and enjoys interacting with them. But friends are also designed to be disposable: to

be an inexhaustible assistant while the player needs them, and to die with maximum pathos

when they don’t. Much work has been done in recent years to make friends more co-operative

(Smith), ideologically biased (Frasca 2001), flexible in their plans (Guckelsberger et al.),

emotionally expressive (Korn et al.) and dynamically complementary in their personalities

(Pacheco and Martinho). But deeper, faster, better friends cannot be created just by making

cuter 3D models or writing ever more tragic backstories into ever snappier dialogue.

Friendship, being a kind of ship, requires a steady ebb and flow. If the player has no duty

towards their purported friend, does not need to reciprocate their care, and has no notion of

their personal boundaries, the possibility of interdependency is eroded.

Briggle writes about the formation of the ethical subject in computer games (165). He argues

that ethical frameworks stem from an acknowledgement of the limits that others place on us.

Because computer game characters are so easily manipulable (172), they are not good

counterparts for a player who wants to treat others well. He argues that “The ethical

169

imperative of respecting others is rooted in the concept of otherness as that which one does

not create or appropriate. Rather, it is received as given—a limit on one’s willfulness” (172). In

other words, proper intersubjectivity requires recognising that others exist and trying to cope

with the fact that you can’t do everything you’d like to them. Better game characters, for

Briggle, would be stubborn, obstinate, would have a certain standard of acceptable player

behaviour, have their own goals and needs, and demand things from the player. Friend

characters, because of their friendliness and the time they spend with the player, thus have a

greater role to play in shaping the player's formation as an ethical being in the game than

other characters.

In Plumwood's terms, game friends are benevolently dominated. Like Briggle, she conceives of

the self as something constructed through relations with others. One kind of selfhood relies on

the creation and subjugation of an opposite otherness. But she proposes another kind of self,

the ‘mutual self’ (142), as a way of being in the world that recognises others as fellows, and

still allows use of them without subjugating them. However, games rarely construct the player

as a mutual self but instead deal in familiar dualisms. By making nonplayer characters

subservient, games make players domineering. This results in disappointing friendships –

because characters are not friends but staff. The ways in which the player is constructed as a

character in the game mirror the ways benevolently dominating selves are made in other

contexts. Because players do not encounter others in games as kin, but as instruments, they

“lack essential (as opposed to accidental) relations to others” (144) and “[experience] the

other solely in terms of [their] own needs for gratification and [their] own desire” (145).

Because nonplayer characters are designed with the needs of the player in mind, they are

othered from the start. They cannot be good friends because, unlike othered humans and

animals, they do not have an independent existence to recognise or commonality to

acknowledge. They are the idealised other, a being that is constructed on the periphery and

waits for the player to occupy the centre.

Othered nonplayer characters construct a certain type of self for the player; they contribute to

the player’s subjectification in the game world. Like subjectification in other contexts, this

performance of the player’s self is iterative and ongoing. Because nonplayer characters inhabit

one half of a dualism, the player finds themselves positioned in the other half. Furthering Yee

170

and Bailenson’s ‘proteus effect’, which posits that players’ in-game avatars influence how

players act in games, and DomíAnguez et al.’s work on players’ automatic adoption of roles, this

conception of the self as relational suggests that if nonplayer characters begin the game

positioned in a certain way, players will be positioned in a complementary way according to

familiar dualisms. If games exist to make the player feel powerful, then nonplayer characters

must be overpowerable (enemies) or disempowered (friends).

For this reason, companion characters are often representations of familiar others: women,

children and animals. Because otherness is something that exists in non-game contexts, the

characteristics of otherness are familiar and already associated with certain types of people.

Companion characters are expected to be feminine, that is, “submissive, accommodating, and

emotionally adept” (Sherrick et al. 19). Brice (2012) notes that when women characters head

a Final Fantasy ensemble they serve as social glue, and “act as emotional guidance for male

characters”. Headleand et al. (2016) observe that players behave differently towards

nonplayer characters whether they are perceived as male, female, or human at all. Curtis notes

that the power fantasy of games is gendered in that “male empowerment depends upon

female disempowerment” (ii). In four lists of the top-ten game companions of all time, sixteen

were female, eleven were animals, ten were male and three were robots (Colyer; Potter;

Sawyer; Wen).

However, companion NPCs are well-liked by some players, and players want companionship in

games (McGee et al). Research also suggests that players like companion nonplayer characters

more than they let on (Merritt and McGee). The development of the earliest digital characters

has some light to shed on their potential for likeability. From 1964 to 1966, Joseph

Weizenbaum developed the chatbot ELIZA, a program which mirrored users’ phrases back to

them in the style of a bored psychotherapist (Weizenbaum 3). Like early computer games,

ELIZA was a demonstration vehicle that could “vividly demonstrate the processing power of a

computer to visitors who did not already have some specialised knowledge” (4). The chatbot

was a programming experiment into the problems (and probable impossibility) of natural

language processing by computers, but Weizenbaum soon found that some users would talk to

ELIZA for hours, often becoming fond and protective of it. That “extremely short exposures to

a relatively simple computer program could induce powerful delusional thinking in quite

171

normal people” (7) was of concern (though not surprise (9)) to the programmer. ELIZA was a

good listener because she was none of the things other humans are: judgemental, blinkered,

expensive, easily bored, or needing consideration and attention of their own. As Crawford

comments, “ELIZA was an empathy-producing machine because she was a simple listener. She

wasn’t trying to be more intelligent than her interlocutors, she was just trying to listen, and

that was actually very powerful.” (Steyerl and Crawford). Her non-humanness was an asset in

the task of getting users to chat freely. Because ELIZA was not human, users felt they could

talk to her without consequence. ELIZA shows that the non-humanness of NPCs is not always

something to be overcome or hidden, but can be an advantage in certain tasks.

Game friends aren’t disappointing because they forget your birthday or ghost on your brunch

date. Game friends are disappointing specifically because they adore you so much, but express

it in awkward ways. Like a cat who shows love by bringing you mice heads, or an

overenthusiastic waiter, game characters try their best to appear friendly but do so in a way

that reveals a fundamental lack of understanding or capacity. Game enemies centre the player

in the ways they spawn, fight and die. Game friends also centre the player, but in an even more

overt way, by going where the player wants to go and doing what the player wants to do. Both

friends and enemies are reactors rather than actors, but because the basis for friendship is

ostensibly the meeting of equals, this unevenness becomes obvious.

FRIENDS OF PHILOSOPHY

As disappointing as they may be to some, to Spinoza, game characters are lovable. To love in a

Spinozan sense is to experience “pleasure accompanied by the idea of an external cause” and

to love results in a person “necessarily endeavour[ing] to have, and to keep present to him, the

object of his love” (140). If a player feels pleasure during the playing of a game, and attributes

this feeling to an NPC, they can be said to love the character. Pleasure, being a transitory phase

in which one’s power to act is increased, is something that friendly NPCs cause all the time.

Friendly NPCs, especially in role-playing games, are designed to be helpful, to give the players

objects and assistance so they can achieve more. But Spinozan concepts of pleasure and love

become difficult when discussing games in which the NPC is the focus of play, such as virtual

172

pet games and dating sims. The NPCs in these games do not really help the player to increase

their power to act, although the game as a whole might. In the satirical dating game Hot Date,

nothing the paramours do helps or hinders the player’s ability to act. The player has the same

dialogue options, the same timing, the same mechanics, the same abilities, no matter what

emotion the NPCs in the game represent or evoke in the player. It is the game as a whole which

is the source of pleasure and the object of love for the player. The collapse of the idea of a

character into the idea of a game reveals that the distinction between the two and the

individuation of a character from its context is something that humans do to characters rather

than something a character fundamentally is.

However, love for nonplayer characters can never be very intense, as “love or hatred towards a

thing which we conceive to be free, must … be greater than if it were felt towards a thing

acting from necessity” (161). This concurs with research suggesting that players care less

about characters they perceive to be controlled by a computer than controlled by another

human (Chesney and Lawson; Fox et al.; Headleand et al. (2015 & 2016); Garau et al.; Lin et al;

Merritt et al. ‘Choosing human team-mates’; Merritt et al. ‘Are artificial team-mates

scapegoats’; Merritt and McGee; McGee et al.). As Khoo and Zubek observe in the context of

enemies, “in the end, the player knows a bot simply does not care, which makes beating it a

thankless task” (48). Love felt towards a computer game character is less than love felt

towards something that is perceived to have free will.

Because enemies in games drop interesting loot and are an interesting combat challenge, they

are also, in a way, pleasurable and lovable to the player. What differentiates the game enemy

from the game friend, in a Spinozan sense, is the player’s idea of causation. Spinoza states that

“in no case do we strive for, wish for, long for, or desire anything, because we deem it to be

good … we deem a thing to be good because we strive for it, wish for it, long for it or desire it”

(137). Good and bad are not intrinsic qualities of an object but rather the result of an

individual’s thinking about an object; a conclusion based on whether that object makes that

individual feel pleasure or pain at that moment. If we take desire to be something that pushes

people towards presumed joyful objects, rather than objects themselves having some joyness

that pulls people toward them (LeBuffe 2.2), then all game characters can be said to be

loveable if the player wants to interact with them in any way. “If emotions are shaped by

173

contact with objects, rather than being caused by objects, then emotions are not simply ‘in’ the

subject or the object”, Ahmed notes (6). Instead, as Gregg and Seigworth observe, “bodies [are]

defined not by an outer skin-envelope or other surface boundary but by their potential to

reciprocate or co-participate in the passages of affect … Affect marks a body's belonging to a

world of encounters” (2). The emotions the player feels towards a character – whether

positive or negative – are a result of the action the player takes towards the character. So the

game enemy is disliked because the player’s goal is to destroy it and the player performs

destructive actions towards it. Because the player seeks to destroy the enemy, and “we

endeavour to … destroy whatsoever we conceive to be [painful]” (149), the enemy is seen as

bad. And so because the friend NPC is something the game directs the player to preserve, and

the player performs affirming and preserving actions towards it, the player feels love towards

it (144).

Interestingly, this Spinozan view of where goodness and badness lies has parallels with how

many NPCs perceive their world. In The Sims franchise, for example, a hungry Sim character

will not decide to go to the fridge to grab a snack, but rather the fridge constantly emits

hunger-satisfying signals, which a Sim will recognise once it becomes hungry. The Sim’s

hunger and satiation is not totally a matter of that Sim’s individual agency, but the feelings are

distributed across game objects and co-created during encounters.

If a person loves an object, they will fear its disappearance and hate anything that could

potentially destroy it: “he who conceives that the object of his love is destroyed will feel pain”

(144); “we endeavour to negative everything, which we conceive to affect painfully … the loved

object” (147). Many games rely on the death of a companion NPC for a climactic event because

the destruction of a loved object is a particularly intensely affective moment. Friends, then,

end up causing the player a lot of fear (fear being “an inconstant pain arising from the idea of

something past or future whereof we to a certain extent doubt the issue” (176)) because their

existence is always threatened. The player loves the friendly NPC, in part, because they are

encouraged to fear their loss. Cuteness and helpfulness are certainly factors in NPCs’ lovability,

but a key component is a game’s mechanics of friendship, the actions the players can take

towards friendly characters. If the player needs to perform friendly actions towards an NPC in

174

order to play the game properly, and is taught to fear their loss, this puts them in the same

place as an object of love.

Players can also pleasurably affect themselves by giving pleasure to friendly NPCs in a sort of

infinite pleasure loop. Giving Lydia from The Elder Scrolls V: Skyrim a fancy glass sword will

make her more effective in battles against the player’s enemies, benefiting the player. Friendly

NPCs then, rather than being pleasurable objects in their own right, are a method through

which the player can positively affect themselves. They can be a conduit of player-directed

affect, even if players discount them as love objects. If a player does recognise Lydia as a sort

of being, though, the act of giving her a sword is pleasurable in itself, as well as a tool for

future pleasure. “If any one has done something which he conceives as affecting other men

pleasurably,” Spinoza observes, “he will be affected by pleasure, accompanied by his idea as a

cause” (150). NPCs aren’t men, but the existence of genres of games based on helping and

caring (like virtual pets, or Animal Crossing), suggests that humans enjoy helping and caring.

Humans will behave towards characters in anticipation of the affect they hope to feel as a

result. They desire certain states. In a study of virtual pet games, Lin et al. found that the

emotions players “desired from having a virtual pet included relaxation, joy, companionship,

and therapy” (4.1).

Spinoza also asserts that “the greater the emotion with which we conceive a loved object to be

affected towards us, the greater will be our complacency” (153) (‘complacency’ in the sense of

security or assurance). This suggests that if an NPC is immediately loving towards the player,

the player will feel secure about their relationship. Because the NPC is so focused on the

player, the player can take the character for granted. Love for NPCs, then, might be best

observed through actions of maintenance and protection, rather than conscious player

emotions.

A Spinozan view of affect gives a general overview of how humans feel about things. Spinoza’s

definition of power is the capacity for action, but does not have much to say about who acts

upon who, or how. Plumwood’s ecofeminist analysis helps to illuminate one of the specificities

of affect, that is, how power shapes human relations to nonhumans. Power is a shaping of

relative affective capacity, such that affect flows different ways depending on one’s position in

175

relation to another. Plumwood suggests that certain kinds of relationships are made possible

and impossible when one party has power over another. Plumwood defines friendship as a

type of mutually sustaining relationship (195). In these kinds of relationships, the self “makes

essential connection [to another, and] fulfils his or her own ends as well as that of the other”

(185). The other “is treated as deserving of concern for its own sake, and hence as intrinsically

worthy or valuable” (185). As far as it can be defined (and Plumwood stresses that it is

dangerous to talk abstractly about structures while ignoring their specific content), friendship

requires “openness to the other, generosity, leaving space for the other, the ability to put

oneself in the place of the other and to respond to the other’s needs” (185). Plumwood

acknowledges that such mutually-sustaining relationships cannot rely on a “personal

conversion to an after-hours religion of earth-worship, tacked on to a basically market-

orientated conception of social and economic life” (186), but the conditions for these

relationships must be made part of wider, public, social and political life. While Plumwood is

talking here about human social institutions, the same principle can be applied to the smaller

worlds of games. Players can certainly feel friendly towards nonplayer characters and can

want to feel certain things, but if the mechanics of the game are such that nonplayer

characters are always instrumentalised, player intention can only go so far in making a true

friendship.

THE FIRST NPC ANYONE EVER LOVED

Most writing about game companions makes its way to Floyd the robot sooner or later

(Frome; Greer ‘digital companions’; Isbister 2016; Rouse; Zagalo et al). Floyd appears in

Planetfall, a 1983 text adventure about having adventures in space and dying in 41 different

ways (Computer Games Magazine). In it, the unnamed player-character crash-lands on a post-

apocalyptic planet and meets a child-like robot called Floyd, who makes a lot of silly jokes and

enthusiastically helps the player solve puzzles. During one of the final puzzles, Floyd sacrifices

himself for the player by retrieving an essential piece of equipment from a room filled with

enemies. In a 1983 review of the game, Creative Computing described Floyd as “the most

imaginative and cleverly written part of the entire game … hysterically funny through most of

176

the adventure, [who] evokes in the player of Planetfall authentic feelings of affection and

attachment.” It further notes that “Indeed, Floyd is critical to finishing the game in terms of

being a large part of the solution to three major problems in Planetfall besides lending his own

brand of moral support to the stranded and baffled adventurer” (Schultz and Arrants). Frome

states that many players cried in reaction to Floyd’s death (158), as does Rouse, and Meretzky.

Zagalo et al. use Floyd as an example of good sadness design, noting that the rupture of a

strong attachment can make people particularly miserable (48).

Planetfall as a game had a few advantages that contributes to the success of Floyd as an

affective character. Planetfall was written by a single person, it was completed within months,

it was one of comparatively few games available for home computers at the time, it followed

closely from a popular science fiction pulp literature tradition, and the idea of an ongoing

friendship with a single NPC was a novel concept to computer game players. These are factors

which are not all possible for games today. Planetfall is also a text adventure about exploring, a

medium and a genre that lend themselves well to experimental gameplay and require some

open-mindedness. To interact with the game, players would also type in their own commands,

which Sali et al. have found is a particularly enjoyable dialogue interface (185).

While exploration text adventures from the 1980s have conventions, mechanics, limitations

and advantages very specific to their kind, Meretzky's decisions in writing Floyd are

applicable to other types of games (Meretzky 2008). Meretzky made Floyd a robot because it

suited the science-fiction setting of the game. The robot seems to be a native inhabitant that

the player happens to stumble across, not a figure placed there deliberately. Floyd's existence

is congruent with his environment; the player can imagine Floyd existing in that world and

living his life happily even if the player had never happened upon him. He is also not human,

and as such, is a bit dim and naturally servile. Meretzky observes that “even with the

concentration on a single NPC, the character would still fall far short of simulating a human

being, but players would have lesser expectations for a nonhuman character such as a robot.”

Making Floyd a robot lowers the player's expectations for the character so that any abilities he

does have are seen as admirable for a robot rather than measly for a human. Although players

expect some human-like characteristics from an NPC (Nass and Moon 82), Meretzky suggests

they have fewer expectations for an NPC that represents something non-human. The more

177

alien and unfamiliar the NPC is, the more inhuman behaviour it can get away with and still be

worthy of consideration.

Floyd could be affected in four ways (Meretzky 2008). Players could issue commands to the

game world, which Floyd would comment on. If the player saved, for example, Floyd remarked

“Oh boy! Are we gonna try something dangerous now?" They could address Floyd directly,

though Meretzky states “there was only enough disk space to handle a fraction of the possible

conversational gambits, so most inputs had to be covered by fairly general defaults”. If the

player asked Floyd to go north, for example, they would receive this in return: “Floyd looks

slightly embarrassed. ‘You know me and my sense of direction.’ Then he looks up at you with

wide, trusting eyes. ‘Tell Floyd a story?’” Floyd could also be triggered to say or do something

by encountering an event or environment. Finally, if the player ended their turn next to Floyd,

Floyd would be triggered to say or do something cute and random, like “Floyd rubs his head

affectionately against your shoulder”. Having established such a cute character, Meretzky

realised “I thought there would be the potential for emotional resonance with the player,

which I could take advantage of by putting this character in peril or even having events lead to

that character's destruction”, affirming Spinoza’s observation on the perils of love (147). Floyd

was useful to the player and assisted them in puzzles, and furthermre reminded them of other

lovable things like children and pets (Meretzky), so his loss depressed the player.

FRIEND AFFECTS AND EFFECTS

Friends are made to affect players positively; they enable player action and make them more

powerful. Things that players do to affect friend nonplayer characters will come back to either

bite or kiss them. How and why players affect friendly nonplayer characters then, is quite

different to how and why they affect enemy nonplayer characters. This section will address

the significant ways in which NPCs are affected, and the mechanisms through which this

happens.

178

Close friends, or companions, follow the player around. This can cause headaches for

programmers and players; friends get stuck in doorways, move too slowly, and generally make

a nuisance of themselves (Botea et al.; Huang and Terzopoulos). It is for this reason that one of

the main problems of friends discussed in games literature is pathfinding.

Friends talk to players more than other characters. But they do not do so in a human-like way.

They can easily become stilted and awkward, repeating the same thing over and over again in

an effort to inform or persuade the player (Domsch 256). They rudely ignore common social

conventions by going on long monologues and not really caring about the player’s responses,

behaving as if conversation were merely an opportunity to take turns speaking rather than a

real exchange of ideas or a way to bond (Fraser et al; Morrison and Martens; Nass et al.;

Smith).

While friends are running around chattering, they are also trying to look at the player-

character. Gazing, and being gazed at, can be an affective action for enemies, friends and

players, or an intensifier for affect. The power of the gaze, though, is unequal. The player can

see the nonplayer character, but the nonplayer character cannot see the player in the same

way.

There are deeper problems with friends than their tendency to wander all over the place and

not listen while staring into the distance. The problem with friends is that they don’t want to

live (Guckelsberger et al.). They want the player to live, but they don’t have any such desire of

their own. Such a being cannot be expected to experience pain, and so their injury or death is

not meaningful to them. When friends fight with the player against an enemy, they go all-in.

Because they have no fear of death, they have no desires, no striving for preservation (Spinoza

137). They rely on the player for meaning. As Emily Short puts it, “Part of what makes NPCs

feel shallow and non-human is their lack of initiative.”

CO-PRESENCE

179

Golland et al. have found that humans tend to synchronise behaviourally and physiologically

when they exist in the same room without directly communicating. When humans interact

with each other they tend to mimic each other’s expressions, voices and emotional states

slightly (Golland et al). This synchronisation tends to increase the more intense the shared

activity is. But Golland et al’s research shows that being silently co-present with another

human also triggers this effect. This is not to say that computer game players will

automatically synchronise their heart-rate with a co-present nonplayer character. But it is to

say that the threshold for affective presence seems to be quite low.

Co-presence, or togetherness, is a technique often employed to facilitate affective relationships

between players and NPCs. Bulu defines co-presence in virtual environments as “Having a

sense of feeling of other individuals, namely perceiving others and having a sense of feeling

that others were actively perceiving us and being part of a group” (155). Co-present NPCs are

continually present within the aural or visual range of the player. They can start off being

friendly or hostile, but over time they become something else: familiar. The co-present NPC

becomes something to be relied upon for a sense of normalcy. Familiarity, especially that

facilitated from time spent together, can also be a first step to more complicated feelings or

connections. Acquaintance can be a road down which other feelings can travel. The design of a

co-present player and NPC is a hope that the player will participate in an orderly and

predictable progression of increasingly intense affection.

In Skyrim, the player acquires Lydia, a bodyguard, who follows the player around and assists

them with fights. The acquisition of Lydia is a reward for completing an important quest,

Dragon Rising, in which the player is tasked with killing a dragon. This quest happens quite

early in the game and is quite hard to avoid if following the main narrative, meaning that Lydia

is often the Skyrim’s player’s first real friend. In the narrative of the game, Lydia is an elite

fighter of sorts, employed by a local king. Being an honourable and chivalrous servant, she is

loyal to, and will fight to the death for, any nobleperson she is assigned to. She lives to serve

and she will die for the player. She is also notoriously hard to keep track of, and will disappear

and pop up in unexpected places. While she is an enthusiastic fighter, she can only attack at

melee range, with the result that she rushes into danger, requires healing frequently, and

180

attacks enemies the player would sometimes rather just run away from. Having her as a co-

present companion, then, affects the way the player plays the game.

Lydia can be romanced, and she can also die. This means she can be attended to in multiple

ways, and requires different things from the player depending on the role the player intends to

inhabit in relation to her. The player can approach her as an employer, a suitor, or as a carer. In

order to keep Lydia alive, the player must supply her with food and potions and monitor her

health. Her health statistics, visible during battle as a green bar above her head (the length

correlating with the character’s wellness), become a concern of the player, another factor to

monitor. If the player is not concerned with the health or feelings of Lydia, they can use her as

a handy portable cupboard (Lydia can carry quite a lot of items).

Lydia’ presence becomes another thing to manage. As well as the player’s own health, capacity,

inventory, weapons, fancy costume, and location, the player must also think of Lydia’s. Her

attachment to the player makes gameplay more complicated and unpredictable, as she runs

into battle or vanishes at an inconvenient moment. But she also increases the capacity of the

player to act in other situations, by carrying stuff and helping quickly end fights. Skyrim with

Lydia is a significantly different experience to Skyrim without.

The mechanism most favoured for the creation and maintenance of co-presence is that of

following - specifically, the following of the player by the NPC as the player moves through the

game environment. But this is not an uncontroversial ability. Guckelsberger et al. have noted

that NPCs can quickly lose the player’s good will and even become annoying if the NPC moves

at a different pace to the player (too quickly or too slowly), blocks the player’s path or has

different goals to the player (1). Co-presence then, is not an easy task to achieve in practice.

Once again, Team Ico shows us how it’s done. In their games Ico, Shadow of the Colossus and

The Last Guardian, companion NPCs (a girl, a horse and a chimera respectively) are always

near, around, or accessible to the player. While these characters follow the player, they do not

do so unconditionally. Yorda (the girl), Agro (the horse) and Trico (the chimera) need

encouragement to follow the player, so the player is continually trying to prove their

trustworthiness to these beings. In this, the following mechanism moves from a subordinate

and locked NPC-towards-player orientation to the continual management of the flow of affect

181

between player and NPC. The player and NPC are constantly beseeching each other. This is

achieved through the imperfection of Yorda, Agro and Trico. These characters are often

obstinate, afraid, wander off on their own, or require a clarifying repetition of commands.

These traits would usually be annoying, but careful handling by Team Ico in the timing and

presentation of these traits serve to make these ostensibly game-ruining characteristics into

expressions of individuality and vulnerability. Tentative deviation from the mechanic results

in a strengthening of its effect. It ensures that the character cannot be entirely known or

controlled, so the player has limited power over it, and a different kind of relationship is

possible.

The game Abzû introduces a variation on the traditional game companion. Although the little

yellow robot that travels alongside the player-character is cute enough and performs useful

functions, its constant presence is not one that inspires affection. Instead, its mechanical form

and bright lights contrast harshly with the soft fluid shapes of the ocean environment, and this

leads to the robot seeming out of place, an interloper that draws the eye away from the

beautiful surroundings. Although the robot is an ally in behaviour, its whirring robotic

presence disturbs the player’s experience of simulated natural wonder. In this case, co-

presence is a factor for annoyance, and the robot’s constant proximity is a constant source of

dismay. As a player friend, this dissonance foreshadows the game’s later revelation of the

player-character as an outsider themselves, a former robotic enemy to the aquatic inhabitants.

The destruction of the companion comes as the player becomes more attuned to their

surroundings, and more distrustful of the machines that threaten the ocean’s wildlife. The

death of this companion is not a tragedy but a release. The inherently uncomfortable nature of

the yellow robot’s friendship serves as a synecdoche to the game’s larger message of

reconciliation with nature.

Co-presence does not have to involve much direct interaction between player and NPC. It

stands in contrast to most other player-NPC relationship mechanics. Co-presence is a passive

action that makes clear the dependencies the player exists within; it requires no action from

the player but has consequences for the actions they can take. There are differences in the

affective power of the NPC depending if they are exclusively audibly co-present, visually co-

present, continually in the mind of the player, or merely omni-present. These differences have

182

to do with power. Nonplayer characters occupy different positions of power in the narrative of

the game and in relation to the player, and this is often evident in the type of co-presence they

share with the player. If a character is co-present but the capacity for one to affect the other is

uneven, this both reflects and reinforces a disparity in power. These power dynamics are often

codified, like in the player’s role as a pet owner or a character’s role as a mentor.

[Image removed due to copyright restrictions]

Fig. 26: The characters of Spider stare at each other. (author screenshot)

The vignette Spider from Vectorpark shows the effect that can be achieved by player

characters and NPCs merely existing in the same space and looking at each other. In this 3D,

third-person game, the player controls a five-legged bird-headed spider which is introduced

alone in a uniform green expanse. Within a few seconds, the player-character is joined by a

much larger character, a six-legged caterpillar-bird-dog, which runs towards the spider

enthusiastically. The speed and size of the bird-dog is initially rather intimidating, and it

seems like the player’s little spider will soon be squashed or eaten. But it soon transpires that

the bird-dog is not a predator but a fan, and only seeks to follow and gaze with fascination at

the player-character. The player can still move their little spider around the scene, but the

bird-dog is always close behind. If the player stops, the bird-dog stops. If the player-character

jumps, the bird-dog jumps back in surprise, but does not stop staring. Whether this fixation is

of love or curiosity is up to the player, though the bird-dog does constantly wag its tail. The

player-character, in turn, stares at the nonplayer character. The triangular beaks of both

characters make it clear what they are staring at. While the characters never touch (the bird-

dog even gingerly lifts its feet to avoid the spider if it gets too close), the effect of the gaze is

still profound. The spider is unquestionably the object of the bird-dog’s attention and while

the effect of being an object of the other’s gaze is intensified by the plain setting, the size

difference of the characters and the implicit predator-prey dynamic association of large birds

and small bugs, the fact of the bird’s constant observation is affecting in itself. To be watched is

the opportunity to affect another, with the intensity of that affect in proportion to the intensity

of the gaze. And that prospect is affective in itself. To be watched, and to watch the watcher

watching, is to be aware of yourself, another, and yourself in another’s eyes.

183

THE GAZE

The gaze is quite an affective event. Fox and Bailenson state that a “gaze can be used to convey

information, regulate interaction, express intimacy, control social interaction, or facilitate task

goals” (150). John Berger defines the gaze as fixed on betweennness, stating “We never look at

just one thing; we are always looking at the relation between things and ourselves” (9). The

gaze is also an important factor in establishing and maintaining power between individuals

(Hoffswell 24; Fox and Bailenson 150; Kleinke). The gaze in games is formed and complicated

by cameras, and analogies to film can be made, especially in terms of the male gaze.

It is the gaze between characters where the differences between first-person games and third-

person games become most apparent. In first-person games friends try to look players in the

eye. But they can only guess where the player actually is and end up staring blankly out of a

screen. In third-person perspectives characters turn and face each other, and the player is left

a voyeur. One of the more interesting developments of gazing in games is the introduction of

VR goggles.

The mouse-like companion Quill of the third-person VR platformer Moss was designed to be

an emotional experience (Lico). She is cute, demanding, and continually looks to the player for

guidance and help as she scurries around a lush forest township solving puzzles. The player in

Moss is a ghost-like point of light who can move heavy objects around for the small animal.

Together, the player and Quill explore this wholesome world. One of Quill’s main strengths is

her awareness of and reactivity to the player. She looks into the player's eyes when asking for

help, after being patted, or if the player is inactive for a certain time. She communicates in sign

language, thus avoiding the disconnect of a sound coming through speakers instead of the

character’s mouth. She waves back to the player, but shows annoyance when the player

demands too much attention. These characteristics are made possible through the VR format

184

of the game, and only enhanced by the player’s status as a ghostly, otherworldy creature

within the game narrative.

Lico states that one of the reasons for the affective success of the companion Quill was the

ability and willingness of the character to hold the steady gaze of the player. Quill has been

designed for maximum cuteness and emotional connection, but one of her more interesting

traits is the ability to look directly at the player. This is not new – virtual pets and companions

of first-person games gaze adoringly out of screens all the time – but the virtual reality

headset’s spatial knowledge of the player’s face and hands lends Quill the ability to look

straight into the player’s eyes. This is a charged event – as Berger remarks, “Soon after we can

see, we are aware that we can also be seen. The eye of the other combines with our own eye to

make it fully credible that we are part of the visible world” (9). Players in games are almost

never seen: it is their avatar who is addressed. The gaze of the nonplayer character is

redirected, or absorbed into an intermediary. To be seen is to be in relation to another.

Other games have used other methods of encouraging players to pay attention when NPCs

talk. Horizon Zero Dawn takes a more cinematic approach, switching cameras back and forth

between focusing on Aloy the player character and her interlocutor, and using fun angles and

the language of cinema to suggest things about the interaction. Conversations between Aloy

and Sylens (a cold, calculating, yet helpful character) are 'shot' looking up at Sylens and

looking down on Aloy, reinforcing the condescending tone of Sylens and his relative power

over Aloy. The camera always focuses close-up on Aloy's face – not just to show off her nice

hair rendering but to indicate that the player is more intimate with her and more interested in

her subtle responses than her conversational companion. Characters do not attempt to look

through the screen at the player: their gaze is turned slightly off-screen to indicate they are

talking to Aloy in the game world, not the player in their world. However, Aloy is still locked in

place until the conversation ends, meaning the player still has a stake in these interactions.

Robert Yang brings male gays’ male gaze into the logic of gaming in his series of games

Radiator 2. The ice-cream consumption simulator Succulent is an interesting case study for the

intense unbroken gaze of its three shirtless protagonists for the duration of its runtime. In

Succulent, a pastel void houses three identical beefcakes: a central figure holding an ice-cream,

185

and two dessertless supporters behind and to either side. The player controls the central

figure’s wrist as he gracelessly fellates a popsicle, looking directly out of the screen. His two

chums also turn to look out of the screen, and encourage the player with appreciative nods as

the popsicle is ground down. This is a confrontational scene, but also an unbalanced one, as

the player is presented with the intense spectacle of a fierce figure aggressively seeking to

please – which the player can also control for the purposes of entertaining themselves, all

under surveillance by the character’s own earnest duplications. The player is confronted by an

NPC begging to be used, but also revelling in the power of being useful.

The spectacle of the powerless using the mechanisms of their objectification to draw attention

to their relative status is not new. In fact, as Steven Shaviro notes in Post Cinematic Affect, it is

a technique that has often been used by the oppressed. By taking the mechanisms of power

and performing them ironically and blatantly, the powerless lays the power relation bare. By

performing usability, the exhibitionist pleasure of being exploited, the popsicle-lover of Yang’s

Succulent reveals the status of NPCs in games as functional exhibitionists, deriving their

pleasure only from giving others pleasure. And the pleasure they give is what Shaviro

proposes is the predominant pleasure of our age – the pleasure of spectacle. Noting the ability

of the performance artist Grace Jones to inhabit animalistic, masculine and savage personae in

order to explode normative conceptions of black women, Shaviro observes the contradiction

of a critical inhabitation of mechanisms as depending on the continuing existence of the

power relations it seeks to destroy. Without the pre-existing implicit agreement that game

assets entirely exist for the benefit of the player (rather than the player being an equal part of

an already-existing whole), Succulent becomes incoherent, or emptied of its critique. Yang uses

the objectified object of the NPC to make a point about queer interpersonal dynamics and

representation of queer characters and queer dynamics in computer games. But in doing this,

he also makes a game about the alternative possibilities of interpersonal relations in games,

and an implicit critique of the kinds of games that can possibly be taken seriously, through his

choice of AAA aesthetic, NPC gaze and role, and queer performance.

DIALOGUE

186

The back-and forth of conversation defines the game friend almost as much as the back-and-

forth of blows defines the game enemy. More has been written about the structures of human-

NPC conversation (Bateman 2021; Domsch; Ellison; E. Short) and how these are technically

achieved (Adams; Campano and Sabouret; Cross; Csurics; Mallon and Lynch; Mikkelson;

Treanor et al. ‘Social Play’) than how different dialogue conventions affect humans or NPCs

(Sali et al). Affect finds its way into the literature eventually though. The improvements

researchers propose for human-NPC conversations are predicated on the specific affect they

aim to achieve. Morrison and Martens propose that “to create richer, more believable

environments for players, we need conversational behaviour to reflect initiative on the part of

the NPCs, including conversations that include multiple NPCs who interact with one another

as well as the player” (1). Campano and Sabouret suggest that “Non Player Characters in video

games should be able to be impolite for a more believable behaviour” (1) and Fraser et al.

advise that their dialogue system “is designed to keep users engaged in the conversation … we

use emotion detection and emotional dialogue management to enhance the conversational

experience” (179). All NPC dialogue actions are triggered by something, but exactly what

affects NPCs and when depends on the type of dialogue system used. This next section will

look at these different systems, and discuss how NPCs are affected by each.

TYPES OF DIALOGUE

Morrison and Martens note that “conversations with non-player characters (NPCs) in games

are typically confined to dialogue between a human player and a virtual agent, where the

conversation is initiated and controlled by the player” (1). Players typically have control over a

conversation (E. Short). The programming of dialogue, the presentation of dialogue, and the

content of dialogue are all centred on the player. These aspects are interdependent and

difficult to tease apart, but are quite different to the NPC. The specifics of a dialogue’s code,

interface and content shape the NPC’s experience of a chat, but broadly speaking there are two

types of dialogue: conversational, where the player and nonplayer character exchange more

than one line, and ‘barks’, that the nonplayer character blurts out after some triggering event.

187

As Emily Short explains, conversation is structured in code based on what designers think

players want out of it. If the player is talking with a nonplayer character about something

important, they may wish to repeatedly return to the topic and review the nonplayer

character’s lines. If the player wants to talk about more than one thing, they may want the

ability to quickly change topic. If the player wants conversation choices to have a meaningful

impact on gameplay, they may find that there are restrictions on the number of things they can

say. These player needs are tied to genre and role conventions, such that a player can come to

expect loquacious companions in an open-world role-playing game and taciturn opponents in

a platformer. These perceived player needs lead to different coding, interfaces and text. First,

let’s look at code.

If the player wants to have a heart-to-heart with an NPC, a common way to organise that long

conversation is through a dialogue tree. There are many kinds of dialogue tree, but a key

similarity is their branching structure. The player is presented with a choice of things to say to

another character, and the ensuing choice leads them down a certain path. At the end of the

path the conversation might be over, or the player might be able to choose another thing to say

to the character. Short notes that “the advantage of a tree-based structure is that it's

(relatively) easy to code, understand, and debug. The disadvantage is that it tends to a certain

kind of design rigidity, and conversation flow is … seldom very realistic”. Branching dialogue

can be quite complicated, with many levels of branches and looped connections. It can contain

conditions, with certain options only being available to the player in certain circumstances. It

can involve dependencies like whether the player-character possesses particular items, has a

high or low reputation, or is running around in the nud (Skyrim). But branching dialogue,

especially in conversations about multiple topics, can still leave the player feeling like they’re

being led down the garden path (Ellison 2), and give an implausible impression of robotic

patience on the part of the nonplayer character (E. Short). In this pseudocode example of a

typical dialogue tree, Short shows how player choices determine the conversational road that

both the player and nonplayer character will follow.

[Image removed due to copyright restrictions]

188

Fig. 27: Short’s example of a typical branching dialogue tree structure (Short ‘Conversation

Design’)

In Short’s code example above, the player is always returned to a ‘hub’, a central place from

which players make decisions. This hub would be visually presented to the player as a set of

options. A set might comprise a list of whole sentences, topics, attitudes, or moods. In Short’s

example, the player would be able to choose from three options: a quip about werewolves, a

quip about wolfsbane, and a quip about Lord Fangclaw. Other topics can be discussed if the

player talks about Lord Fangclaw first. The code also shows that once a topic has been

exhausted, the player can choose to talk about something different. In conversations where

the player can talk to a character about a number of topics, it is difficult to find natural segues

between subjects. In reference to hub-based dialogue structures, Ellison comments “this

method of dialogue tends to create conversations strongly divorced from reality. The NPC

usually has infinite patience for the player's strange inquisitions, and every dialogue plays out

like an interrogation as the player keeps pressing the NPC for info. Furthermore, the player

hears a lot of the same lines over and over as he navigates between hubs” (3).

Other kinds of conversation structures exist to solve these problems. A simple branching

dialogue in which neither the player nor the NPC are given a choice about their lines allows for

a better conversation flow, though at the expense of player choice. Matleuf compares his

experience navigating Tales of Monkey Island, which uses hubbed dialogue trees, to Secret of

Monkey Island, which doesn’t. In Tales of Monkey Island:

“You'd choose from a list of dialogue options, each one branching off into a handful of

further options. Once all those were exhausted it'd boot back to the previous menu.

Rinse, lather, repeat until everything that could be said would be. This enabled the

player to see all the conversational content without having to reload a previous save or

replay the game. … it hurts the proceedings in a variety of ways. For one, conversations

would take too long. Each episode in Tales of Monkey Island starts with a good dose of

exposition with little interaction as the player clicks their way through dialogue. …

More importantly, it takes the fun out of having to carefully choose what to say, since

you know you'll eventually be able to say everything anyway.”

189

Secret of Monkey Island had a different dialogue system. In this game, the player can only

choose one thing to say to an NPC, but this choice has no impact on the outcome of the game at

all. Players are free to choose the silliest option and receive a silly reply in return. Matleuf says

of Secret of Monkey Island, “No longer could you just keep picking the top option, confident

that you'd eventually click your way through all the zippy punchlines. You had to stop and

think what you wanted to say as the comeback would probably be pretty funny and something

you wouldn't want to miss out on.”

Short suggests both conversation flow and player choice can be maintained by using an atomic

model of dialogue. In this coding structure, dialogue options are presented one or two at a

time to the player, and have prerequisites for showing up. These prerequisites might be based

on how long the conversation has been going, what has been already been said, the mood of

the NPC, or details from the surrounding world.

Other kinds of speech besides back-and-forth conversation exist in games, and these are a

little easier to program. ‘Barks’ are short, innocuous phrases that friendly characters utter

when a player gets too close. The character might be thinking out loud to themselves or

greeting the player, but the ostensible effect is for “breathing life into a space, to really bring it

alive for a bustling, living feeling” and providing a “small but intimate glimpse into the lives of

[a game’s] inhabitants” (Mikkelson). In Skyrim, the content of barks can be affected by the time

of day, what the player-character is wearing, what they’re holding, if they’re sneaking, if

they’ve dropped an item, if they’ve taken an item, or if they’re in the NPC’s personal space.

Outside of conversation, NPC speech can be triggered by the usual things NPCs are triggered

by: proximity of the player-character, scripted events, changes to the NPC’s body, the player-

character’s body, or the world, the beginning or the end of a level, time, and location, though

the exact things that affect the production of speech and the content of speech depend on the

individual game. Like proximity can provoke an enemy into attacking, proximity can provoke a

friend into confessing whatever’s on their mind.

Skyrim offers one good example of contextual barks. If the player-character unexpectedly

enters their house nonplayer characters will ask the player “You lost?", "Can I help you?" or “I

190

trust you're not planning any trouble. What can I do for you, friend?” Enemy characters cry out

one of twenty-two variations of “That's it? That's all you've got?" when the player hits them.

Depending on their race and mood, townspeople will say something like "Hey now, you be

careful with that fire" if they notice the player carrying an open flame. If a nonplayer character

is a vampire outside during an eclipse they will say "The tyranny of the sun is over!" (A nice

list of these one-liners and their conditions can be found at the Unofficial Elder Scrolls wiki.)

In some games, NPCs chat amongst themselves (M.U.G.E.N doesn’t offer competitive AI-

controlled chatting as an option unfortunately). In Skyrim, pub patrons tell each other about

their knee injuries as the player-character walks in. Stealth games like the Assassin’s Creed

series depict dialogue between nonplayer characters, either to create a sense of realistic

camaraderie between characters or to give a sneaking player the thrill of juicy eavesdropped

gossip (Domsch 256). In Final Fantasy VI characters whisper secrets to each other, even

though the player controls most of them and all text is presented in text boxes (the boxes show

phonetic whispering sounds).

After all that, there’s cutscenes. Action games sometimes implement cutscenes, during which

the player-character and nonplayer characters act out an important piece of exposition, but

the player is not in control. These scenes are like little films, though Short notes they are

usually one of the few places where a nonplayer character can appear to take the initiative.

HUMAN SPEECH ACTIONS

BjoH rk and Brusk have analysed dialogue structures in games for how they exhibit common

design choices in light of BjoH rk, Holopainen and Lankoski’s work on categorising ‘gameplay

design patterns’ (Brusk and BjoH rk; BjoH rk and Holopainen; Lankoski and BjoH rk 2007). Patterns

here refer to “reusable gameplay design choices as semiformalized and interconnected

descriptions that may turn design possibilities into explicit options” (Brusk and BjoH rk 1).

Lankoski and BjoH rk build on BjoH rk and Holopainen’s player-centred work in a series of papers

on designing more engaging and believable nonplayer characters (Lankoski; Lankoski and

BjoH rk 2007; Lankoski and Bjork 2008; Lankoski and HelioH ; Lankoski et al.). Following on from

this, Brusk and BjoH rk show how these gameplay design patterns apply to game dialogue in

191

particular. However, their way of categorising patterns is not particularly intuitive and often

mixes player actions, game actions, player character actions, and emergent things. They

identify a few player conversational actions:

• Single Initiative Dialogue: the player acts and the game reacts; the game does not take

initiative (3)

• Interruptible Actions: a player acts to interrupt an NPC (2)

• Canned Text Responses: a player acts to further dialogue along a dialogue tree (3)

• Free Text Communication: the player can type whatever they like and the game will try

to understand it (3)

• Cut Scene: the player acts to initiate a cut-scene (3)

• Location-Specific Dialogue: the player takes conversational actions in a specific location

(3)

• Character-Specific Dialogue: the player takes conversational actions with a specific

character (3)

• Context-Specific Dialogue: the player takes conversational actions in a specific context

(3)

• Relation-Dependent Dialogue: the player takes conversational actions having formed a

specific relationship with a character (3)

• Affective Communication: the player acts to modify the NPC’s mood or disposition

towards the player (3)

This obviously isn’t an exhaustive list. The dearth of academic and game design texts about

player conversational actions and their associated interfaces suggests that it is an

underappreciated area. Finding analyses of dialogue from the nonplayer character’s point of

view is even more of a challenge. With that in mind, let’s start at the beginning.

The first thing players do to affect nonplayer characters through dialogue is to initiate it. For

non-conversational speech they can perform one of the bark-triggering actions as described in

the previous section. Starting up a conversation might not be a deliberate action either. The

player might just find themselves automatically talking to a guard in The Elder Scrolls IV:

192

Oblivion if they’re caught stealing a nirnroot. But when players choose to converse, by clicking

on or walking up to a character, they do so through an interface.

Instead of acting through the player character’s body, players usually speak through an

exegetic intermediary in the HUD which is invisible to the game inhabitants and evident only

in its effects. That games have dialogue interfaces that are different from other kinds of

interaction can be affective to the player in itself. Short notes “When all conversation takes

place in its own interaction mode and is separate from the other action of the game (which

stops for the duration), the NPCs come to seem as though they primarily inhabit some other

plane of existence from the player character”. Different interfaces means players face different

choices to make in a conversation. The player’s part in the conversation might involve typing

complete sentences into a text parser, speaking into a speech parser, choosing from a list of

something. Sometimes the player’s role is just to mash buttons to prompt characters into

exchanging predetermined lines. Or the player’s only choice might be whether to get a cup a

tea while they wait out a cut-scene.

Sali et al. found that “holding the rest of the game steady, changing the dialogue interface

produces significant changes in gameplay experience. Further, these changes shape

perceptions of the system well beyond the interface and its operation” (179). They compared

three different dialogue interfaces: natural language input (the player types whatever they like

and the game tries to understand what they mean), dialogue menus that specified the exact

words the player-character would say, and dialogue menus that presented some abstract

approximation of what the player-character might say (showing a shortened version, a topic,

an attitude, or an emotion, for example). Sali et al. found that while abstract responses were

preferred by designers, they felt less conversationally natural to players, and made them

speculate on the mechanics of the game. Full-sentence responses, on the other hand, were

disliked by designers who felt like they were too long to read and created distance between

players and their avatars, but liked by players who weren’t bothered by reading time and felt

more engaged in the narrative of the game. Natural language inputs were considered a huge

technical headache by designers, and somewhat error-prone, and made players feel that they

had less control over the game than the first two dialogue interfaces. However, players enjoyed

and engaged with the natural user input more than the other options. Sali et al. conclude that

193

“game dialogue interfaces have a profound impact on the experience of gameplay” (186). As

Short notes, “The world of gaming dialogue is not quite divided into ‘Façade’ and ‘everything

else’”, though it has to be said that there is a lot more scripted dialogue in games than there are

natural language parsers.

In Red Dead Redemption 2, dialogue options are contextual. The player has access to different

speech acts depending on the circumstances they’re in. Dialogue options are presented on the

lower right hand of the screen, alongside other interpersonal actions. There are usually two to

five options, which the player can select with a click of a mouse or a press of a controller

button. The options are generalisations of tone: typically ‘aim weapon’, ‘rob’, ‘greet’, and

‘antagonise’, with the name of the nonplayer character underneath.

[Image removed due to copyright restrictions]

Fig. 28: The player can cheer or heckle Aldridge T Abbington in Red Dead Redemption 2.

(author screenshot)

Once the player chooses one of these options, the player character says something

appropriately hostile or cheery, in his own words. Subtitles appear if the character says a long

sentence, but don’t always appear for short phrases.

Both the player-character’s and nonplayer character’s words might be conveyed through

audible speech, subtitles, a text box, or a combination. The voice actors might be speaking the

player’s language, an unfamiliar language, or gibberish (or, as in Animal Crossing, familiar-

sounding, localised gibberish). Voices and text might be constructed out of words spliced

together (Adams) or reproduced in full. Though it has fallen out of fashion in recent games, a

trend for a while was the silent protagonist. In an effort to immerse the player in a blank slate

of an avatar, some games’ player-characters were mute, either to the player or to both the

player and the game world (Domsch 253). As Miozzi points out, this sometimes results in

bizarre scenes where all the nonplayer characters in the game are happily chatting away while

the player-character stands around awkwardly.

194

A particularly interesting example of the silent protagonist and of weird dialogue interfaces is

the speech-craft wheel from Oblivion. Speech-craft is a skill in this game, and if players get

good at it they have a higher reputation among NPCs and can charm their way into or out of

sticky situations. Oblivion has a pretty standard conversation system, with players able to

choose one-word topics from a list to prompt an NPC to riff on. But at the bottom of the

dialogue menu is a small square with a face on it. Choosing this leads to a kind of mini-game of

abstract repartee. In this rather bizarre interaction, players quickly and wordlessly flatter,

threaten, impress and humour an NPC in quick succession. The NPC offers some short verbal

quips and brief facial expressions during this process. If, after this, the NPC is sufficiently

buttered up, they might tell the player extra information, give them a better price, or let them

off the hook for a crime.

[Image removed due to copyright restrictions]

Fig. 29: A cropped image of the persuasive speech-craft minigame from The Elder Scrolls IV:

Oblivion. (author screenshot)

It’s hard to avoid awkwardness in conversations with NPCs, especially with interfaces like

these. As a result, many modern games let the player skip through dialogue, dramatically

decreasing the amount of time players have to stand around listening to an nonplayer

character. Talking to an NPC isn't really like talking to a human or an animal. It's not even

really like talking to Siri or Alexa. Players have a pretty clear idea how conversations will turn

out (Treanor et al.), and normal rules of politeness are out the window (Campano and

Sabouret). Domsch further points out the disadvantages of verbose nonplayer characters. He

recalls the phrase “I used to be an adventurer like you, then I took an arrow in the knee” from

the game Skyrim (256). This was a throwaway phrase, something a guard would say to the

player as they walked by, or a pub patron would say to introduce themselves. However,

Skyrim’s casual dialogue repository was so small that it quickly seemed to the player that

every third townsperson was a lame ex-swashbuckler. Instead of indicating individual

quirkiness, the line of dialogue emphasised the generic souls of Skyrim’s population.

195

The lack of player choice can also be affective. When the player controls a silent protagonist or

a taciturn player-character this limitation can bolster the impression of the player-character

as their own being with their own way of doing things. The player might like to ask Sylens of

Horizon Zero Dawn about the ethics of human-chicken relationships in video games, but Aloy

isn't interested, and so the option isn’t presented to the player. In some circumstances, like

dating sims, the limitations of dialogue seem like the boundaries of polite conversation, so that

the player's choices are constricted by decorum rather than the decisions of a game dialogue

writer.

NONHUMAN SPEECH REACTIONS

The effect of dialogue on NPCs is profound. Players’ speech is powerful, having an effect on

NPCs like a law or force. The player's dialogue choices can change the course of a game, and

can change the life of an NPC. In Skyrim, for example, one sentence from the player

determines whether Lydia becomes an adventuring companion or sits around forever in an

empty castle. Lydia only starts following the player if the player chooses the dialogue option

“Follow me. I need your help.” Once the player has said this, she is bound to the player,

following them everywhere and battling creatures on their behalf, until the player tells her to

leave.

For a game about yelling at lizards, the player’s voice seems to have a powerful effect on

Skyrim’s townsfolk too. In an example of conditional dialogue, the player can woo and wed

characters if they have an Amulet of Mara in their inventory and the nonplayer character is

positively disposed towards them. These player attributes affect the nonplayer character’s

conversation options before the player even speaks to them and affords further affection. If

the player owns the amulet, proposing out of the blue becomes an available dialogue menu

item in every conversation with a friendly nonplayer character. Sweethearts (other than

Serana) will either say yes immediately or send the player on a simple quest before saying yes.

After that, they will be waiting on the steps of the temple the next day. If the player finds

themselves busy and doesn’t show, they just reschedule the wedding. The marriage ceremony

is similarly straightforward, with both the player-character and the nonplayer character

196

declaring their commitment. This speech act has permanent effects; divorce is not an option in

Skyrim. Marriage changes a newlywed NPC’s affective capacity. They can be ordered to do

chores, they can be asked for money, and they confer oblique benefits on the player (in a

beautiful example of game logic, if the player-character goes to sleep in their house with their

spouse they receive a 15% boost to skill-learning unless they are a werewolf). The nonplayer

character is affected by marriage in ways other than domestic servitude, however. If the bride

or groom was already a companion of the player-character, they become immortal. They move

into the player’s house, if the player has one, or share their house with the player if they don’t.

If this house is in a different city, they might change job too. They can create meals without the

aid of cooking equipment. Speech acts, then, can uproot a nonplayer character’s life.

Short’s code above focuses on the actions players take. There are a number of ways NPC

usually respond to these quips. Each of the player’s statements might have a corresponding

line written for the NPC, so the character just respond to the player’s call. The NPC might have

a set of answers from which they randomly choose. The NPC might secretly check the player-

character’s pockets and look them up and down to see whether the player has the right

requirements to hear one answer or another. Or the NPC might do a combination of these

things, choosing a random answer after the player satisfies their requirements, or adding a

random wisecrack after a scripted line.

Nonplayer characters also change themselves by speaking. They mostly do this by taking

secret notes during conversations and remembering past conversations. If a player has

already talked to a nonplayer character before, they might refuse to speak or give a short

rebuff. Nonplayer characters can also register information about the player-character and

modify their future speech based on their observations. Hades is an exemplary example of this.

In Hades, repeating the same levels over and over again is the whole point of the game, and

nonplayer characters are in on the joke. As well as commenting on the player’s achievements

and items, NPCs recognise the player-character and say something a little different every time

they die and come back. The player can even act as a go-between for two separated characters,

advancing the conversation through their rotation through levels.

[Image removed due to copyright restrictions]

197

Fig. 30: A nonplayer character in Hades recognises the player-character, admiring them for

playing the game repeatedly. (author screenshot)

[Image removed due to copyright restrictions]

Fig. 31: The player insults a stranger for no reason in Red Dead Redemption 2. (author

screenshot)

Nonplayer characters also have good memories in Red Dead Redemption 2. In Figure 31 above,

after the player has insulted the stranger, the stranger runs away from the player-character,

yelling conciliatory remarks. If the player ever encounters the stranger again, he will be wary,

or even run away from the player before the player has a chance to call him a tiny mushroom

brain again.

Nonplayer characters can also invite themselves to parties or change their orientation to the

player during conversation. In the Oblivion example above, the player works to improve the

nonplayer character’s disposition towards them, making them more or less friendly. The Final

Fantasy series is infamous for nonplayer character betrayal, where characters who appear

friendly turn into enemies halfway through a conversation. In the first example of this, a king

in Final Fantasy asks the player to collect his lost crown from a cave, a very ordinary friendly

request. When the player returns with the magical item, the king transforms into a demon and

fights the player. Initiating conversation with the king switches him from a friend to an enemy.

G. Smith also observes that “Final Fantasy VII sometimes switches insiders and outsiders as

the context changes”. In Final Fantasy games, in the process of talking, nonplayer characters

join the player’s team without the player being consulted at all. Conversation results in a

character changing from an ordinary NPC into a kind of player-character themselves.

Nonplayer characters’ conversation actions can change the animation of the character. A facial

animation might be linked to a certain point in the conversation, the character might

demonstrate an action, or the character might transform into a new shape. Animation in

games is a huge topic, but what is is important for this thesis is that it can be prompted to

198

change, no matter the mechanics of what happens afterwards. Characters in games behave like

Spinoza’s stone – they keep doing something until they’re made to do something else. The

animation of their 3D models or the substitution of their 2D sprites is part of an overall

behaviour. It can be thought of as an appropriate outfit a nonplayer character wears for an

occasion. If a player in Oblivion plays the speech-craft minigame with a nonplayer character,

for example (see Figure 29), landing on the “joke” quadrant will, all things being equal, prompt

a smile from the nonplayer character. Things being unequal, like the reputation of the player-

character and the nonplayer character’s disposition towards them, will prompt the nonplayer

character to grimace or show a blank expression. Facial animation has moved on since

Oblivion, but the fact of its situational prompting remains.

The most significant way an NPC can affect the player is to give the player a task. Nonplayer

characters can “literally tell the player what to do!” (Mikkelsen) through quests, hints and

advice. Warpefelt’s typology gives further insight into this behaviour. Quest-givers “often are

there to provide the player with a diegetic interface to quest management, to provide the

player with the mission, and then to reward the player when they have completed the

mission” (38). In open-world role-playing-games like Skyrim, ‘radiant story’ AI means NPCs

never stop literally telling players what to do. As Delahunty-Light puts it, “Played all those

Dark Brotherhood assassination contracts that never seem to end? Trawled through Vex’s

limitless supply of thieving jobs? … Radiant story produces those kinds of simple, repeated

fetch quests, designed to ensure that you always have something to do and with enough

variation … to keep you from repeating exactly the same quest over and over.”

When an NPC proposes a quest to a player, they are inviting the player to participate in an

affective exchange with them. In Skyrim, when a player agrees to a quest, that quest is added

to a journal, and markers appear on their map. In a quest called “Grin and Bear it”, for

example, a character called Temba Wide-Arm expresses her hatred of bears, and asks that the

player kill ten bears and bring the skins back to her. The player can go out and hunt for the ten

bears or just buy ten bear pelts from a shop. If the player returns to Temba with the skins,

Temba will give them an enchanted weapon. If Temba dies after that, the player will also find

that she has left them a small inheritance. This is a pretty basic quest, but it serves as a simple

example of the kind of affective arrangements players can get themselves into.

199

The content of NPC speech obviously affects players, or at least it’s supposed to. Characters’

style and content of speech is tied tightly to their function. A game’s quantity and quality of

speech is linked to genre. Domsch observes that the more emphasis a game puts on narrative,

the more dialogue it includes (251). Adventure games, for example, are often set in

complicated fantasy worlds that require a lot of explaining. But adventure games also promise

adventuring, and players want to get to the action. In the Final Fantasy series, this tension is

resolved by short, snappy sections of characters’ colourful storytelling. G. Smith observes that

“because players are frequently impatient with character speech in games, sometimes seeing

these spoken interludes as obstructions that keep them from more interesting battle

scenarios, the dialogue in games needs to be particularly pithy and efficient.” However, he

argues that dialogue in games is ‘freer” than that of film or television because games do not

have to rely just on dialogue to move their plots forward. Instead, periodic enemy battles

provide a sense of urgency and action. Interpersonal drama can be driven just as effectively

through blows as with barbs. Dialogue in games, then, is free to fulfil functions other than plot-

moving, and is often used for worldbuilding, backstory, and giving a sense of history.

Mikkelson explains that such dialogue makes the player “feel really involved and immersed in

the world as opposed to just an onlooker” [emphasis in original]. This conception of dialogue

as a method of telling the player about the world and their purpose within it is common, but

necessarily positions the player as an outsider. They are just an onlooker until they are invited

into the world by a friendly local.

Conversations, especially in older games, are big events, as both the player and the nonplayer

character stop what they are doing, turn to face each other square on, talk without fiddling or

performing any other action, and often do it at a cinematic camera angle. When a player

decides to converse with a character in The Elder Scrolls IV: Oblivion, the camera zooms in on

the character's face, so they are large and central in the player's vision. This mimics the effect

of attentiveness in that the character is really the only thing the player can see, but it's a

clunky approach that reproduces the effects of being affected rather than producing the

desired affect. It also makes interaction with nonplayer characters seem both separate from

the rest of the game, and more intimate.

200

[Image removed due to copyright restrictions]

Fig. 32: An example of a dialogue interface from The Elder Scrolls IV: Oblivion. (author

screenshot)

In Oblivion, conversation topics appear in the bottom left-hand side of the screen. The player

can choose from any number of these topics, and the NPC will respond patiently to each

abrupt change of subject. Most Oblivion residents are massive gossips, so ‘rumors’ is almost

always an option. The player character does not have voiced lines, and their speech is not

subtitled. The player is left to guess exactly what their player-character would say and how

they would say it.

HOT DATE

[Image removed due to copyright restrictions]

Fig. 33: The interlocutor of Hot Date is unimpressed. (author screenshot)

Romance games, visual novels and text adventures offer players the greatest opportunity to

affect NPCs through dialogue choice. Hot Date is a dating simulation game about the timely

seduction of self-involved pug dogs. The game is set during a speed-dating session in a dingy,

dim cafe. The player is presented with a talking dog and has a limited but unknown amount of

time to get to know them. These dogs all use the same 3D model but have different body

language and facial expressions, which gives them distinct personalities. They start the

conversation by introducing themselves. While the player cannot choose their interlocutors,

look around, or walk away from the table, they can choose questions of varying silliness from a

menu on the top-left of the screen, and type in nouns when asked to by their date. Actions are

limited to the player choosing when to speak, selecting set questions from a list (albeit rather

a large list), and typing in responses as prompted. The game is controlled by the mouse and

keyboard, but these do not power movement of the player character’s body, rather movement

of the conversation. Engagement is maintained by the unpredictable emotional reactions of

201

the dog, and the acerbic nature of its comments. These emotions are shown in the body

language of the pug, which exhibits human-like reactions, and the content of its comments.

The player can monitor these signs for a sense of how the conversation is going, and so tailor

their questioning approach accordingly. The pug has an opinion of the player on a spectrum

from “extreme dislike” to “like”. The player can influence this through their choice of question.

The aim of the game is unclear. A happy romance is not possible, as all the player's potential

dates are weirdos who will cease to exist after a few minutes. The player can certainly try to

make each dog love or hate them to the greatest extent possible. The player can try to deeply

understand each individual dog they come across. Or the player can just try to get through all

the funny dialogue options. It is difficult to maintain the dogs’ interest or avoid offending

them. Even if the player does somehow get a dog to like them, there is no future for the

relationship as the paramour is soon replaced with a new surly suitor.

In Hot Date, the player has the odd experience of interacting with a fictional being who has

completely written them off. In games terminology, the dog is no longer immersed in the

game. The player is not worth bothering with; the player is less affective. This is actually a

rather liberating experience. The player can interact with the pug as a nonplayer character in a

game instead of keeping up the fiction of speed-dating. After a few dates with these

judgemental dogs, it is less of a temptation to discover the inner lives of the animals than to

see how far they can be pushed in one direction or another. The player’s list of predefined,

surreal questions also prompts the player to start playing flippantly. This emotional distancing

is not only a result of the enormous amounts of insults the player has to endure from

successive sanctimonious paramours, but also a desire to experiment with the game itself. The

game invites frivolous play; the dogs obviously aren’t taking the situation seriously either. It

becomes a pursuit of affect for the sake of affect, the player acting on the dog just to see what

happens, the dogs disengaged in turn. Through this, the game moves from the player playing

with the NPC, to the player playing with the rules of the game itself. The game become less of a

situation in which an NPC’s psyche can be explored, than one in which the total affordances of

the games’ programming can be revealed.

202

Dating games ask the player to form a non-normative romantic attachment that cannot be

fulfilled in conventional ways. Hot Date preemptively scuttles any chance of love by presenting

the pugs as horrible people, so the player is also off the hook for not being a genuine suitor.

Neither the player not the NPCs are suitable partners, and so in the context of romance they

arrive on equal terms. The awkward mechanics of small talk serves as its own reward.

FEELINGS

NONHUMAN FEELINGS

To a friendly NPC, a human player is a repetitive, demanding, myopic character who moves at a

frustratingly slow and inconsistent pace. From an NPC's point of view, the human is the one

who is boring and nonsensical. The player’s existence compels the NPC into new ways of

being, but the player-character is also bossy, incredibly slow, keeps getting distracted with

invisible things and can’t perceive obvious things. The roles of humans and NPCs determine

what they can be to each other, and this not only limits the potential design of game

characters, but makes the possible human response to those characters similarly robotic. It's

fun to act as a leader, making all the important decisions and setting others the work you'd

prefer not to do. But the result of this is that the player becomes a boss, othered from their

NPC allies not only by their differences in material embodiment and conscious awareness, but

in their place in the world. Plumwood remarks that hyperseparation in labour “is often

framed in a mind/body dualism in which mind people control body people” (207). She further

points out that the idea of a naturally servile, mindless class is a very old concept (84). While

the role of the player is that of a leader who makes decisions and sets tasks, the experience of

the nonplayer character consists of responding immediately when called. But the player of a

game is also subject to the calls of the game’s design, forming a loop of affective power. As

Anable puts it, “Video games ask us to make choices, and they ask us to operate within the sets

of constraints or rules that govern those choices” (xii). Hammer and Baker further this point

in saying “In a computer game, these rules can’t even be negotiated with; they’re enforced by

code” (2).

203

NPCs feel drawn to humans. The appearance of a human is an exciting event in an NPCs

existence. NPCs are launched into new behaviours, compelled into new animations, activated

by new functions, and otherwise oriented like a compass to the player’s north. The NPC's

experience of the player is that of a god (reflected in the word ‘avatar’). The NPC discovers

capacities they didn’t know they had. Their brief existence before the appearance of the player

seems idle. They can only react, even if the player is unaware of what they are reacting to.

These directions are mostly internal, within the NPC’s code, although the NPC has no

preference whether commands come from orders or urges. However, some NPCs have choices,

and some degree of randomness. Their programming may give them the ability to react

differently to the exact same stimulus. What NPCs feel in that case, is what Nash would call

autonomic aleatoricism (20), or a pre-individual dependence on chance.

Friends’ representation of emotions can affect the player’s ability to act in the game. Happy

friends, Isbister observes, cheer the player along and lend moral support (228). Friends also

express fear and disagreement. In Ico, as the player guides Yorda through the castle by the

hand, she occasionally hangs back and shakes her head, afraid of a wobbly surface or

vertiginous height. Yorda is weaker than the player and needs constant assistance. She helps

the player, but she also needs help from the player. Although the player controls only the body

of Ico, Yorda and Ico are a team, and the game can only be completed if the player makes

Yorda's skills and needs central to their strategy. This mechanic also features in The Last

Guardian. The game makes it clear that the boy is the main, playable character, and Trico is a

mysterious, independent other. But the player really controls both characters, one directly

through buttons, and one indirectly through the boy’s suggestions. That The Last Guardian’s

controls were criticised for being unreliable and glitchy only emphasises that players of the

game were controlling a relationship rather than an individual.

THE LAST GUARDIAN

204

The Last Guardian is a game about learning how to work with huge unwieldy frustrating

beasts. In it, a small boy and a large chimera must solve puzzles together to escape a

dilapidated castle. The player plays as the unnamed boy, who affects Trico through voice

commands. The boy is small and clumsy and Trico is the size of a small elephant. Both

characters start the game injured and wary. It is clear that in the world of the game, these two

are traditional enemies. Trico needs to be befriended and fed before it will follow the player,

but it does not rely on these treats and pats for survival. Instead, the player relies on it to

navigate the oversized landscape. Like other NPCs, the beast is player-oriented, but what

makes The Last Guardian and other virtual pet games interesting is that the win-state of the

game requires the player to be NPC-oriented. The player must take interest in the state of the

NPC’s body and mind in order to progress in the game. As in most virtual pet games, the player

character is invulnerable while the pet itself is prey to all kinds of ailments and afflictions. The

aim of the player is to maintain the animal’s body and shape its behaviour.

There is no HUD (heads up display) in The Last Guardian. There are no weapons menus or

inventory. This is a cinematic game in which everything the player can use is an object present

in the world. There is no pausing the game to check on supplies or to switch between rifles;

there (apparently) is nothing the player can see or access that Trico can’t. This makes the

game seem less like a game and more like a situation.

The first task of the player is to earn Trico’s trust. From the beginning of the game it is clear

that the player’s focus must be the beast. He clearly needs help – he is chained to the ground,

is impaled by spears, and hungry. Given that there is nothing else to do, and Trico is the large

grey thing in the room, it falls upon the player to assist. By tentatively approaching Trico (not

quickly, not near his face), the player-character can seize one of the spears and remove it. The

player has the option to run away at Trico’s howls, or to pat the animal in its distress.

Investigating further, the player learns that they can grab onto the animal. It can be climbed, it

can be ridden. Trico is still wary, and still seems like a dangerous creature. There are barrels

scattered around the level. Barrels, as all game players know, contain valuable loot. But these

barrels cannot be opened by the player-character. After moving the barrels within reach of

Trico, the player finds Trico crunches it up (the barrel is as big as the player character himself)

in one go, demonstrating intimidating destructive power. The player-character cannot be

205

damaged in the traditional game sense, and neither can Trico. However, they can present

distress, and this, rather than health bars or quantified stats, is the player’s motivation for

keeping both happy. The game at this point is entirely centred around the emotions of an NPC.

The player and Trico move through the castle haltingly. As well as the abilities of walking,

climbing, jumping and grabbing, the player also has the ability to call Trico and pat Trico.

These abilities are mapped to buttons on the player’s game-control device, indicating their

importance to the gameplay. The player, being small, continually has to crawl through tunnels

or shimmy along ledges to unblock some gate or activate a mechanism to clear a path. With

the player-character on his back, Trico can jump long distances, move huge obstacles, and

blast things with his electric tail. It is clear that neither could escape the castle on their own.

Trico’s design is friendly, but also truly chimeric. His body is covered in grey feathers and he

has four pink bird feet. In shape he is somewhat like a griffin, though his face is like that of a

dog or possum. He has small wings, which allow him to jump and glide, and small, broken,

blue horns on his head. He has a long tail which fluffs out at the end. From the size of his wings

and horns, his uncertain demeanour, his skinny form, and the proportions of his face, he seems

young. Trico moves nervously but lithely, like a cat. He is pretty damn cute. A chimera in

behaviour as well as form, he communicates in dog-like, cat-like, bird-like and NPC-like ways.

His messages are subtle and the player must play close attention to his body language.

This chimeric design helps make Trico attractive to a wide range of people. Spinoza notes that

“simply from the fact we can conceive, that a given object has some point of resemblance with

another object which is wont to affect the mind pleasurably or painfully, although the point of

resemblance be not the efficient cause of the said emotion, we shall still regard the first-

named object with love or hate” (141). So when Trico the digital chimera sits back and

scratches his ear with his back leg, it reminds the player of their own pet cat doing the same

thing, and although neither the pixels on the screen nor the specific act of scratching is

inherently lovable to the player, the image is pleasurable from its association with an existing

pleasurable thing, and so the player goes ‘aww’. The resemblance carries the affect over from a

memory into the present situation.

206

This association, however, can also lead to unfavourable comparison. In a study of how players

felt about the dogs of a VR virtual pet game, Lin et al. investigated how emotional connection

with real pets affected their expectation for virtual pets. It turns out humans compare virtual

animals to real animals constantly, and this affects their perception of the virtual animal. Two

particularly disappointing aspects of virtual pets were “(i) the dog's reaction to external

stimulation, such as fetching or petting, and (ii) its behavior patterns, such as walking or

running.” (5.1.2).

Trico judges the player silently. It is unclear which player actions influence his opinions. It is

clear that some counters are counting and some variables are varying, but there is nothing

other than Trico’s emotional display to tell the player they are doing well. The player can pat

Trico, feed him and command him, but he offers no immediate response, just stores it in his

memory. The player’s actions are meaningful to the character, but the player does not know

how. The chimera’s inner world is a mystery, but unlike other game characters, an intriguing

mystery. In reviews of the game players noted that Trico doesn’t like being ordered around too

much and spamming the controls seems to make Trico ignore the player (emily_). Whether

Trico’s unpredictable behaviour is subtle and realistic or the result of a terrible AI is a source

of much fan discussion. There is a lot about the game that is clunky. In an interview, the lead

designer Ueda seems to suggest that this inconsistency was both intentional and difficult to

maintain during development, saying “You might call Trico to come over and he doesn’t come

over. Since I’ve been working on this in development for years upon years, I’ll go ‘c’mon, just

come over.’ [laughs] I get irritated. Then, I have to remember: if I’m a player who’s never

played this game before and they call Trico over and he doesn’t respond, that’s the more

realistic response. Trico’s an animal that doesn’t follow logic, the logic that we humans would,

so it’s actually a positive thing. That’s the thing that I always need to keep in mind. The longer

you play, the more logical you tend to try and make things, and that’s something we have to

resist [as a developer]” (Klepek). The Last Guardian is often criticised for having unsteady

controls and being too finicky to handle. The cameras often end up at weird angles, the player-

character doesn’t reliably respond to controls and Trico doesn’t always do what he’s told.

These flaws were particularly frustrating when the player had to act quickly or precisely, but

they also added verisimilitude to the child character. The boy seems clumsy, tired, and not

quite comprehending what he’s supposed to be doing. He seems out of his depth. These

207

inadequate beings who are both flawed in the game world and flawed as game characters

demonstrate that better friends can be made out of worse friends.

HUMAN FEELINGS

From the amount of research that exists into making game friends better, humans seem to feel

disappointed but hopeful about game friends. No doubt someone enjoys game companions,

but in the game design literature (including this thesis), NPCs are positioned as a problem to

be solved. Most work into human feelings towards game friends, then, emphasise player

dissatisfaction in order to posit a solution to it. Human feelings towards NPCs might be

broadly categorised through the lens of game design literature into successful feelings

(feelings that designers want players to have) and unsuccessful feelings (feelings that players

don’t want players to have).

NPCs make humans feel awkward. In designing NPC behaviour in Skyrim, awkward behaviours

indicated areas of NPCs’ AI that needed to be revised. “When every time I drop something on

the ground, half the town comes over, it is starting to get really silly” the lead designer

Nesmith, interviewed by Lenhardt, remarks. In a study of virtual pets, Lin et al. found that

players felt weird about virtual dogs. This feeling, however, was connected to different aspects

of the virtual animal. “We found that participants commonly attributed their negative reaction

to the appearance of the virtual animals, which could elicit feelings of disgust or fear.

Conversely, positive emotions were most often attributed to the behavior of the virtual

animals. For instance, some participants cited that the way the virtual animals ran or

responded to lights made the animals seem fun.” (Lin et al. 5.1.2)

As Lin et al. show, the appearance and behaviour of an NPC has some bearing on what players

feel toward them. Headleand et al. observe that the represented species and gender of an NPC

elicits different behaviours from players, which may indicate different feelings. Role and

player expectation also influence how humans are affected by NPCs. Guckelsberger et al. state

that “in a qualitative study on companion behaviour, a player said ‘I dislike that [the

companion] prioritises getting to the exit herself over helping [me] first’, stressing the delicate

208

balance between support and independence” (1). Warpefelt (2015) argues that if designers

want players to feel something specific or particularly strongly, they must make sure that

“NPCs must act in accordance with the expectations of the players … it is of the utmost

importance that NPCs are designed in such a way that they look feel sound and behave in

accordance with what is signalled to the player” (1).

When designers want players to feel awful deliberately, they make the player lose something.

“Many video game players and designers consider making players cry the ultimate proof of a

game’s emotional effectiveness” Frome states (156). Zagalo et al. further point out that “The

videogames industry has tried to find solutions to this problem of interactive sadness for a

long time” (45). Both Zagalo and Frome conclude that when humans cry about games, they

usually cry about the unfortunate fates of their sidekicks, who have mostly died sacrificing

themselves for the player. Zagalo et al. don’t take a strictly Spinozan approach to sadness, but

come to the same approximate understanding of sadness as a passive, low-energy state. They

observe that it’s hard to induce such a feeling of disempowerment in a medium about action

and control, and note that most of the more celebrated sad moments in games involve

cutscenes or reduced interactivity (46).

BETTER, FASTER, STRONGER FRIENDS

Making game friends better is a thriving area of research. Much work focuses on how to make

game friends more believable, with methods like realistic behaviour, subtle emoting, emoting

more subtly and frequently, becoming more aware of and reactive to their immediate physical

and social context, and of course, navigating doorways.

Grace notes, “the relationship of mechanics in games is like relationships, also complicated”

(42). Many approaches have been taken in creating affective friends. One common approach

has been to create stable personalities for NPCs (Arrabales et al; Bates; Georgeson and Child;

Khoo and Zubek; T. Short). These researchers posit that a distinct personality not only forms

209

the foundation for an NPC to consistently and idiosyncratically act and react to the world

around it, but makes it more human-like (and therefore better) in the eye of the player.

According to Higino et al., Kasap et al. and Verhagen et al,, the meticulous construction of a

good solid identity provides the foundation for an NPC’s dynamic and naturalistic behaviour.

Nonplayer characters, to Lankoski and BjoH rk, are part of a “designers’ means to control their

players and guide their gameplay” (2007 416). They agree with most other scholars that

believability is the key to making nonplayer characters more affective (416) and argue that in

order to be considered believable humans, nonplayer characters in games should acquire

some “Awareness of Surrounding, Visual Body Damage, Dissectible Bodies, Initiative, Own

Agenda, Sense of Self, Emotional Attachment, Contextual Conversational Responses, and Goal-

Driven Personal Development” (416).

NPCs work best when they are alone with the player. In a group, they do not engage in the kind

of communal dynamics that humans find natural and comfortable. They seem awkward and

gruff around their digital fellows. NPCs’ focus on the player character as a source for meaning

and action comes at the expense of other social relationships. It’s notable when NPCs interact

with each other, whether it’s the player coming across a fight between NPCs in Skyrim, a fight

between dinosaurs in Ark: Survival Evolved, or a fight between cowboys in Red Dead

Redemption 2. In a mirroring of the origins of human-NPC relationships, it seems the initial

way that NPCs are beginning to interact with each other (unscripted) is through violence.

Interest has grown in the ability of NPCs to work together with each other and together with

humans in group tasks (Abraham and McGee; Afonso and Prada; Bailey and Katchabaw; Dimas

and Prada; Gruenwoldt et al.; Higino et al.; Jansen; Verhagen et al.). Gruenwoldt et al. note that

“one thing that modern video games still lack … is a sense of relationship or social network

binding the characters and objects in the game world to one another” (2), and Jansen observes

“In most current computer games … artificial intelligences do not really cooperate with their

companion. They co-exist beside their partner, doing their own tasks, without attacking the

partner player” (1.1) while Afonso and Prada remark that this is something players

desperately want. NPCs have close relationships with the player, but distant relationships with

210

everything else around them. In order to be pleasing to the player, then, they should care less

about pleasing the player and focus more on having well-rounded lives.

Some of these researchers have specific tasks they would like nonplayer characters to be able

to perform, and identify deficiencies in current design practices. Verhagen et al. explain that

“to enable NPCs to display behavior consistent with our expectations of human player

behavior, and to be able to form teams of NPCs as well as hybrid teams consisting of humans

and NPCs, the internal NPC model needs to change” (1). In order for players to happily and

productively work in mixed teams of humans and nonplayer characters, nonplayer characters

need to change to become more human-like, they argue. Verhagen et al. further note that most

nonplayer characters have no awareness of their social context, and so can only act based on

their physical environment and programmed desires. Knowledge of “status, roles, norms in

place etc.” (1) is important for getting along with humans well.

Nonplayer characters are single-minded beings who have problems reading rooms and going

with flows. This, Dimas and Prada, argue, is because the identities of NPCs are rigid and fixed,

in contrast to the more fluid, situated, and dynamic identities of humans. If NPCs are to be

richer characters, they posit, they must also become more contextual and malleable people. If

they are to act naturalistically and believably in a group setting (whether a group of NPCs,

humans, or a mix of the two), they must take the lead from humans and be self-conscious

wrecks like the rest of us. This, of course, requires that they are aware of their context in the

first place.

When humans encounter other humans, they search for telltale expressions of an essential

inner character (Isbister 5; West and Zimmerman 129). They recognise "something that can

be conveyed fleetingly in any social situation and yet something that strikes at the most basic

characterisation of the individual” (Goffman, quoted in West and Zimmerman, 129). Small,

subtle mannerisms build up into indications of a whole personality type. What humans see

when they encounter others is a performance of a deep, stable identity. But what nonplayer

characters express to others about their essential natures is a complete lack of it. Their

essential nature isn’t something intrinsic to them, but something constructed for them for a

purpose. “A character becomes a game character once it is integrated in the game’s mechanical

211

system which requires the player’s non-trivial effort to progress from one state to another” as

Blom puts it (10). Players evaluating nonplayer characters must take into account both the

character’s absence of an inner self, and their performative expression of that self. Nonplayer

characters are in a tricky spot when it comes to being recognised as individuals with their own

unique personalities, because what is fundamental about them is their artificiality. Not only

can nonplayer characters not express an essential nature, most players won’t even look for

one, instead assuming qualities of their characters from their relative role or their context

(Warpefelt 2015). Players encountering nonplayer characters know that they aren’t real

people. The more nonplayer characters try to appear to be people, the more they fail. When

NPCs try to do humanness, they always appear to be falling short or overcompensating.

NPCs would be better people, McDonnell et al., Korn et al., and Kasap et al. argue, if they

reacted more immediately and intensely to their environment. NPCs often don’t react to

events within normal human ranges of time and expression, instead offering delayed and

underwrought performances. These unconventional reactions can be a source of comedy,

especially when an NPC appears nonchalant or oblivious of unusual things happening around

it. Making NPCs more dramatic involves researching what makes humans so dramatic and

trying to replicate the basics during game development. Korn et al. have a good overview of

this process and compare the different emotion models game designers use in their work

designing emotionally expressive rock trolls. Suffice to say, trying to figure out how human

emotion works and then programming that model into an NPC to make them affectable in

certain ways and so express represented feelings properly is a tricky process.

Many researchers talk as if the pinnacle of nonplayer character design would be making a

character indistinguishable from a human, in behaviour if not in appearance. While that

certainly would be cool, it is also worth thinking about ways in which nonplayer characters

could be better without them having to be more human-like. Nonplayer characters are already

quite good at lots of things – being frustrating, being persistent, being efficient, being silly,

being predictable – and although these qualities are the result of convention and technological

limitation, they can also be said to be the nature of the beast. Nonplayer characters don’t need

to be very technologically sophisticated to perform the sorts of tasks people require of them

(Khoo and Zubek), and much of what is necessary can be achieved with good art rather than

212

novel programming. Much of what makes Trico and the lizards of Rain World appealing is their

good looks and smooth moves. And because these characters are not trying to be human, they

cannot fail at being human; they can succeed at being weird cool game characters rather than

trying to reach Turing test standards. This view also positions NPCs as something designers

do to players, as tools for affect, which, as Brice (2017) notes, requires players to “hop into a

dark space and keep our fingers crossed that the other person knows what they’re doing”

(80).

As Johnson and Verdicchio observe in the case of social robots, “It is not that we have

discovered a new entity and are poking and prodding it to figure out what it is. Rather we are

at once creating a new type of entity and simultaneously asking what it is. This process is

especially complicated in the case of humanoid robots because of their human-like

characteristics and the trend towards making them more and more human-like” (292). Trends

towards making nonplayer characters more or less human-like in any way is a decision

humans make. But particular humans make them for particular reasons. One frequently cited

reason is market forces. Johansson notes that game designers think players both cannot

accurately evaluate increases in NPC complexity and aren’t that interested in it anyway (206),

and so putting in extra effort to make the best NPCs ever probably wouldn’t result in the game

developers being especially popular or financially successful afterwards. NPCs don’t even need

to be improved at all if their marketing is good enough, or links are made between advancing

computer abilities and all-over gaming goodness. As Rouse says in reaction to the over-hyped

promises of the PlayStation 2’s processing abilities, “it seems that calling the PlayStation 2’s

enhanced graphics chip-set the emotion engine is most likely a buzzword some clever person

in the marketing department dreamed up rather than anything an intelligent game designer

would use to describe such technology” (Rouse 6).

Not to propose another model of human emotion for games, but let’s talk about Spinoza. NPCs

are affected, and they are affected in ways specific to their kind. To embrace a Spinozan game

design ethic would be to recognise NPCs not as a kind of human-lite, but as something that is

always individuating within a constantly changing and relational context, and working to

enhance those individuating tendencies. A good example of Spinozan design is Vectorpark’s

work Spider, discussed above, and Ball, discussed in the next chapter. Spinoza could also be

213

used in an attempt to make NPCs more affectable in a human way. A Spinozan view of affect

does not require actors to have solid identities or even solid bodies in order to be affected. For

a human, to be affected is to form associations between unrelated things that happen to occur

at the same time, or in the same place. Through memory, these associations build up over

time. Keeping in mind that people can be affected by imagined or remembered objects just as

intensely as anything else (143), this means that people exist within idiosyncratic networks of

meaning, constantly affected not only by things around them, but what those things remind

them of. As Mateas says about the potential of AI, the advantages of more sophisticated AI that

can observe and associate are that it could generate more “behaviour that a player can read

meaning into” (1).

To make an NPC more like a Spinozan subject, then, means making them remember. If an NPC

is to be affected like a human, this memory must not only store information about specific

events, objects, and people, but document co-present events, objects and people, and create

connections between them. Kasap. et al. have looked at implementing memory in NPCs for the

purposes of better emotional expression in interactions with humans. Their model creates an

emotion engine for NPCs that feeds events through a simplified personality which affects how

the NPC reacts: “the virtual character’s personality … alters her appraisal of the event” (23).

This engine includes a memory for emotional events, meaning that an NPC who experiences

something positive or negative will carry that impression over to similar events. The NPC that

Kasap et al. has designed then, can form an impression of a human player, can modify that

impression over time, and can react to the human player based on that appraisal.

Where Kasap’s NPC experiment innovates is not only in sophisticated emotion processing, but

its positioning of the player. When NPCs are programmed, they are programmed with a certain

interaction and a certain kind of player in mind. The designers intend for the player to act a

certain way; they have an ideal player (Aarseth 2007; Sicart 2008) in mind who plays exactly

how the designers would like them to, and seamlessly and frictionlessly experiences the

designer’s pure intentions. Kasap et al.’s NPC does not have a preference for how players will

behave toward it. The player can act any way they like and the NPC will react to that specific

behaviour, rather than the assumed action of the ideal player. NPCs are not usually

judgemental people. They doggedly stick to their task no matter what is going on around

214

them. They don’t have the ability to play along when the player is acting silly. Their rigidity

means they provide a predictable and reliable experience, but only if players follow a

complementary script to the NPC. By making their NPC judgemental, Kasap et al. has

paradoxically freed players to behave as they like.

However, this dynamic is specific to an academic, experimental setting. In other digital

environments, it might not be so charming. Making NPCs more human-like might have

unexpected drawbacks besides its impossibility. The social awkwardness of NPCs can be a

factor in their likeability. As the thousands of Oblivion memes show, NPCs can be appreciated

for their unique weirdness. Much pleasure can be had in watching Skyrim’s dragon get tangled

in a tree and explode (oOFrostByteOo), killing, cooking and eating a baby dinosaur an inch

away from its emotionless mother in Monster Hunter (Sterling 20:30), or following random

oblivious people around for the whole day in Red Dead Redemption 2 (DefendTheHouse). The

Sims franchise is arguably entirely built on these strange player-NPC interactions (Martey and

Stromer-Galley; McGuire). By aiming to make NPCs less like uncool weird robots, designers

may have the unintended effect of making NPCs less like the cool weird robots they actually

are.

Such a framework always sees NPCs as fundamentally lacking: they are not human and so they

are less than human (except in the rare case where they are the evolutionary precursor to

some apocalyptic AI, see Whitby and Coeckelbergh for further discussion on this). If the

alterity of the NPC were accepted, new design directions could be found that both increased

the complexity and believability of the NPC without making them compete against humans for

humanness. This does not even have to result in any adjustments to existing conceptions of

personhood (though that’s probably a good idea in any case).

OBLIVION

215

The capabilities of Oblivion’s now-kneecapped NPC AI system are almost mythical at this

point. Aerothorn, bitmob, Delahunty-Light, Hughes, Lenhardt, and Mullon all recount the story

of how Oblivion’s Radiant AI system was so powerful, and made NPCs so lifelike that it had to

be diminished so the self-willed characters didn’t run amok and ruin the game world for the

player. Radiant AI was a system developed to make the NPCs of Oblivion more complex and

lifelike. It gave each NPC a daily schedule to follow, gave them strong motivations, memories,

personalities, and attachments, and kept track of their interactions with the player and each

other. It was an attempt at procedural behaviour. Oblivion’s NPCs get out of bed in the

morning, sit at their breakfast tables, go to their workplaces and perform work animations,

head down to the pub at lunchtime, meet with their friends for a weird chat in the afternoon,

sit at their dinner tables, and get into bed in the evening. They don’t just stand around in one

spot waiting for the player to find them, they have their own simulated thing going on. But

something went wrong with Radiant AI. It became too powerful, and NPCs started behaving in

strange ways. Intelligent ways. Ways that would not be advantageous to the player. In an oft-

quoted example of the full power of Radiant AI, game designer Emil Pagliarulo recalls:

In one Dark Brotherhood quest, you can meet up with this shady merchant who sells

skooma [an addictive drug]. During testing, the NPC would be dead when the player got

to him. Why? NPCs from the local skooma den were trying to get their fix, didn't have

any skooma, and were killing the merchant to get it! (bitmob)

Radiant AI promised NPCs that lived their own lives and behaved in spontaneous,

idiosyncratic ways. It aimed to create a functioning world of simulated townspeople for whom

the player was not the centre of their lives. But it is unclear even fifteen year after Oblivion’s

release whether Bethesda really did hobble the world’s best NPC AI system because it was just

too good, or whether it was all just marketing hype. Bethesda has never released the code for

its Radiant AI, so no one has not been able to check if it could actually do what its designers

claimed (see Aerothorn, Hughes, Hacker News, and the two Wikipedia talk pages for some

heated forum discussions on this). The beautiful potential of Radiant AI can only be found

through anecdotes and spurious unsourced claims.

216

In a paragraph that has been copied and pasted around the internet so many times an original

source cannot be found, Oblivion’s Radiant AI apparently resulted in these strange and

unacceptable scenarios:

The following are examples of unexpected behavior discovered during early testing:

• One character was given a rake and the goal "rake leaves"; another was given a

broom and the goal "sweep paths," and this worked smoothly. Then they

swapped the items, so that the raker was given a broom and the sweeper was

given the rake. In the end, one of them killed the other so he could get the

proper item.

• Another test had an on-duty NPC guard become hungry. The guard went into the

forest to hunt for food. The other guards also left to arrest the truant guard,

leaving the town unprotected. The villager NPCs then looted all of the shops,

due to the lack of law enforcement.

• In another test a minotaur was given a task of protecting a unicorn. However,

the minotaur repeatedly tried to kill the unicorn because he was set to be an

aggressive creature.

• While testing to confirm that the physics models for a magical item known as

the "Skull of Corruption," which creates an evil copy of the character/monster it

is used on, were working properly, a tester dropped the item on the ground. An

NPC immediately picked it up and used it on the player character, creating a

copy of him that proceeded to kill every NPC in sight.

This particular copy-and-paste is from Wikipedia’s The Elder Scrolls IV: Oblivion talk page

second archive. While these anecdotes are widely used to describe the effects of Oblivion’s

fully-powered Radiant AI system before it was toned down, it cannot be verified. When bits of

this text appear in academic articles, they also reference forum posts that have no original

source. These behaviours are obviously very cool, and some of them could possibly

understandably justify reprogramming a game. Some of these seem like normal bugs, some of

them seem like interesting emergent behaviour, but some of them seem very unlikely.

217

The NPCs of Oblivion, then, are always compared to the fully-realised, self-actualised beings

they could have been. They are not only a disappointment compared to real humans, but also

to their idealised former selves. As Robert Mullon pointed out in a review of Oblivion on

AlteredGamer,

Upon playing the game the player will notice that an archer isn’t just another

“cardboard cutout” but actively prowls the forest to hunt a deer; similarly, a citizen

stops after work-hours and retires to a favourite inn. In principle this should go some

way towards the creation of a more immersive and more believable environment;

however, in practice its execution really seems nothing more than regurgitated

heuristic-algorithms with a polished marketing name. The Radiant AI system was

rarely commended mostly due to the fact that actors behaved the exact opposite way

intended. Rather than give each NPC a distinct personality, the new AI system

effectively created a world of clones following the same schedules and actions.

The history of games marketing is the history of disappointing game releases. Oblivion’s NPCs

are a joke now not only because of their surreal glitches and relatable awkwardness, but

because these flaws point to a failed ambition. They create an image of the kinds of people

they tried, and failed to be. What is not addressed in lamentations of Oblivion’s Eden of AI is

why it was ‘dimmed’ in the first place. The consensus among commentators and game

journalists seems to be that it was changed to make the social world of the game neater and

more comprehensible for the player, and so important characters didn’t get killed before the

player reached them (Aerothorn; bitmob; Dafaure; Hughes; Mullon;). Whether Oblivion did

have a fantastic AI and removed it for player comfort can’t be known from current available

sources, but it’s interesting to note how readily accepted this story is. The dimming of

Oblivion’s AI is a tale about how player comfort and comprehension is, and should be, centred,

even when it leads to bad NPCs. And conveniently, the problems with Oblivion’s NPCs all come

about because the developers are just too good at their jobs.

Oblivion’s NPCs really are just bizarre sometimes. They do affect players, but not in the way

the designers intended. Compilations of their weird speech and behaviour have become a

popular topic on YouTube as quintessential examples of NPC strangeness. The reprogramming

218

of their AI gives an excuse for this behaviour. Apocryphal tales from Radiant AI present it as an

Eden of NPCs, and the reprogramming as a downfall. Radiant AI probably wasn’t as good as its

legend has it, but the idea of what NPCs could become remains embedded as part of Oblivion’s

experience.

CONCLUSION

Affective friends in games have been a fertile topic of research because they are a challenge to

design well. The perceived requirements of friendship means that friends present unique

challenges that make them particularly difficult characters to create. Designing friendship

often relies on the cutting edge of AI technology, and so the failures of game companions

reveals the limits of game technology. Good game companions have to be acutely aware of the

player and react to them quickly and appropriately. Even when the game companion isn’t

trying to simulate a human this is a difficult ask. But when game friends are human-shaped,

expectations are even higher. Creating believable characters has been a goal of games research

for some time. But what is believable is not only a technical challenge, but a challenge to

imagination. Game characters will never be perfect simulations of humans, so the task is to

make players care about them in spite of that fact (or even to appreciate them for their non-

humanness).

It is more difficult to design a friend than an enemy. Apart from the issue of doorways, friends

are asked to do more complicated tasks. Because more is required of them, they have more

opportunity to fail. Friends are tasked with simulating friendliness. One way designers try to

achieve this is to make characters visually appealing according to conventional

understandings of attractiveness. Another is to design the mechanics of the game so the player

and the NPC rely on each other to fulfil the player’s goals. In this model of friendship, NPCs

offer the player resources, and the NPC and player spend a lot of time together. Through co-

presence in an emerging narrative, the player and the NPC hopefully develop as characters

together. Because all aspects of friendship cannot be included in a game, the actions designers

219

choose to include reveal both what they think are the essentials of friendship, and what

computers are capable of in terms of simulating human behaviours.

Serious attempts have been made in the pursuit of good game friends. Oblivion's Radiant AI,

whether its wilder stories can be believed or not, was a real effort in making more

independent NPCs who lived in a self-contained world. Oblivion’s designers had the goal of

creating a world in which the player was a part, not the whole. Its NPCs were given the

semblance of self-determination for the purpose of creating a real ecology and decentering the

player. These characters are the butt of jokes now (and were at the time), both as a result of

game development decisions like cutting corners on dialogue writing and voice acting, but

also because giving more freedom to a computer program to act does not always result in

naturalistic human behaviour. While many point to the dimming down of Oblivion’s AI as the

source of its NPCs’ clunkiness, their awkwardness more likely comes from the fact that they

tried to appear human-like in the first place.

Trico of The Last Guardian and Floyd of Planetfall are two of the more successful

developments in game friendship. There are a few factors to their success. Floyd succeeds for

a few elements not relating to his design: the text-based, narrative driven, sci-fi inspired

nature of his game, and his novelty at the time. But there are commonalities between Trico

and Floyd that point to a possible way forward for designers wanting to make good, affective

companions. They are both non-human characters. The player relies on them to get them

through the environment, and cannot act without them, Floyd is human-like in his ability to

communicate, and is also human-shaped. But his robotic body, and Trico’s animal body, mean

that the player expects robot-like and animal-like behaviour from them, rather than human-

like behaviour. The perception that animals and robots are less capable than humans works in

Trico and Floyd’s favour, as players are pleased with what the characters can do, rather than

disappointed at what they can’t. The two characters are also constantly by the player’s side,

and the player relies on their unique abilities to move through the game’s landscape and plot.

They are also vulnerable, requiring the player’s attention and care. And their narrative was

tragic – two innocent beings caught up in difficult circumstances, who respectively die and are

exiled, for the benefit of the player at the climax of the game.

220

Sophisticated technology isn’t required for good game friendships, but it does help. And it also

helps if designers design backwards from what they hope to achieve from friendship rather

than trying to make friends fit into a world made for player characters (i.e. if NPCs keep

getting stuck in doorways, design worlds with bigger doorways). Friendship has to be taken

into account when designing the levels and mechanics of the world, otherwise the relationship

feels unnecessary. And players cannot be friends with NPCs at all if they are unwanted.

221

CONCLUSION

NPCS EXIST

Discussions of the problems of NPCs’ affective capabilities have overshadowed the realities of

how humans and NPCs actually affect each other in games. Though the fuss about the ethics of

human-NPC relationships seems to have died down a bit in the past decade or so, the majority

of literature concerning them is still focused on how such interactions subsequently affect

human relationships, whether in a positive or negative way. Furthermore, where affect is

discussed, it is usually in terms of a lack or excess, and this is located in the NPC itself rather

than the interplay of NPC and human. NPCs are either too affective or not enough, and both

are a problem to be solved. Either way, they are important. “The most important dimension for

storytelling in games is that of agents/characters. This indicates that the most effective way of

creating ludo-narrative content is to invest in character-creation, by making the characters

rich, deep and interesting.” Aarseth says (132). This thesis continues the tradition of finding

fault with NPCs by acknowledging that they often aren’t that great, but identifies the solution

to this problem in the structure of the game world rather than the superfluous or deficient

qualities of the character or the player. It also comes to an ambivalent stance towards the

affective capacity of NPCs, acknowledging the affective work they do perform and seeing their

underwhelming or odd impact as part of their charm.

This thesis came from a dissatisfaction with NPCs that wasn’t rooted in any mechanical or

representational failure on the part of the characters, but in the ways players were called to

perceive and act towards them. It tries to achieve a few (too many) things. Firstly, it argues

that decentering the player will result in more affective human-NPC relationships. Secondly, it

tries to achieve this decentering in a small way itself by describing the ways in which NPCs

affect and are affected when a player plays a game as intended. Thirdly, it uses an ecofeminist

lens to explain how power differences between subjects are formed and maintained in games

the same way they are in other contexts. Overall, this thesis hopes to be one possible answer

222

to the vague question ‘what if we took NPCs seriously on their own terms’. This conclusion will

give a quick overview of these arguments, but primarily it will look to certain games,

especially small indie games, to see what can be done in terms of making relationships

between humans and NPCs weirder and cooler, in an effort to achieve the aims stated above.

NPCs are blamed for a lot of problems in games and the world, and if NPCs had any sense at all

they would point their digits right back. NPCs aren’t failing so much as they are being failed.

This seems strange to say about fictional things with no existence prior to their fabrication by

humans, but as long as NPCs are made as useful things and not as affective beings, they will

not be able to be as interesting and realistic as they could be, and as everyone wants them to

be. This thesis comes down pretty firmly on the side that instrumentalist relations in games

are a concentrated and simplified reflection or reinforcement of a pre-existing ideology rather

than a creator of them. In other words, NPCs aren’t inherently bad, but the ways players treat

them mirrors how the powerful treat the powerless in other contexts. Little evidence suggests

that NPCs have significant effects on humans, but quite a lot of evidence suggests that humans

significantly affect NPCs.

That this thesis relied mostly on free, cheap, discounted and borrowed games that could run

on old PCs speaks to both the huge numbers of indie games made by people who receive little

to no money from them, and their importance to advancing computer games as a medium.

That small indie games can be essential, widely recognised and badly compensated says

something about success in the games industry and the intractability of a lot of its problems.

When larger games experiment with new ways to relate to others, like The Elder Scrolls’

Radiant AI and Radiant Story, old relationship models somehow seem to become even more

entrenched, albeit in a new procedural way. Games need new relationship models, and need to

experiment with non-normative ways of being with others. It is for this reason that queer and

independent games form the basis for this conclusion’s optimism.

Nonplayer characters are not player-characters. They have a clunky name, but it’s a perfectly

accurate one. It reveals the complete interdependency of the player-character and the

nonplayer character. Like the self and the other, the player-character and the nonplayer

character cannot exist without each other. A nonplayer character cannot exist without a

223

player-character (Ball being an arguable exception), and if we take a nonplayer character to be

a concentrated embodiment of a game’s affective intentions, the player cannot either. That

might be a bit of a fudge, but NPCs are inextricable from their context and attempts to

individuate them are doomed. They are not separable from their game environment, they

cannot survive outside the game they are in. Their code is the game’s code, their behaviour is

shaped by rules and prompted by external events, their body is a temporary arrangement of

impersonal dots. NPCs are not different from humans or other animals in their ontological

indefiniteness. But as M.U.G.E.N., merchandise and fan fiction show (and Blom expertly

details), game characters are also ideas and exist in the minds of humans. Like other groups

subject to power, they are defined from the outside. But because they don’t have an

independent self to refer back to, this puts their affective capacity in a weird spot.

Games can be said to operate in an affective economy in which particular affective states and

intensities are valued, and certain exchanges are normal to make. In a Spinozan sense, nature

is the only individual around, and the rest of us poor sods are modes, territories (to use a

Deleuzean term), or instances (to use a games programming term) who are not separate from

our affective relations or “plane of consistency” (environment or background) (Deleuze ii).

This means that computer games and their constituent parts, as things that exist, are just as

part of (Spinoza’s capital-N) Nature as anything else, and are as affecting and affectable as

anything else. How affect works in digital spaces then, is not so different from any other space,

though game-ness is its own affect or its own specific place (as discussed by Huizinga in his

conception of the magic circle, itself critiqued by Consalvo). Chang reminds us that in games,

“emotion becomes part of a player’s apparatus for negotiating virtual environments” (75);

“emotions are not secondary to the base functionality of an organism, but are instead vital,

innate cognitive tools that it uses to navigate and react to its environment.” (75).

Since game characters don’t have a body, a soul, human-like self-awareness, a continuous

existence, or an ability to freely communicate, their ability to define themselves is limited. As

an unconscious object, their lot in life is to be described by humans. And so the question of

whether players or game designers have the final authority over the true identity of a game

character is unfortunately inevitable, but still interesting. This problem was exemplified by a

series of articles about RimWorld, a game which found itself at the centre of a small

224

controversy in 2016. RimWorld is a third-person top-down management-simulation game

about a group of stranded astronauts building and maintaining a small society on a hostile

planet. In it, the god-like player must keep characters alive, build and maintain a habitable

environment, and combat outsiders, by assigning tasks to the astronauts. The game was softly

released in 2013, with many updates and patches added until its official release in 2018. In

2016, the developer added the ability for characters to socialise with each other, and the

journalist Claudia Lo looked at the source code of RimWorld to find out why straight male

characters couldn’t stop driving themselves into depression (Lo). It turned out that straight

male characters would automatically attempt to romance any attractive woman, no matter

their paramour’s sexual orientation or if they had already been rejected. Their perpetual

fruitless wooing of attractive lesbians resulted in the male characters’ moods perpetually

decreasing, which had an effect on the kinds of actions they could undertake and the social

dynamic of the space colony (RimWorld is programmed according to Spinozan conceptions of

pleasure and pain). But in looking at the code, Lo found other interesting drivers of

RimWorld’s character behaviour. Straight women and bisexual men did not exist in the game:

all female characters were either bisexual or homosexual, and all male characters were either

heterosexual or homosexual. Male characters only found female characters between the ages

of 20 to 40 attractive, but female characters only found older male characters attractive. Being

repeatedly romantically rejected affected the male characters’ mood, but being repeatedly hit

on did not affect the female characters’ mood. Male characters always hit on characters they

found attractive, but female characters only hit on characters they found attractive about 12%

of the time. As Lo noted, “this scenario mirrors a common narrative about romance, sexuality,

and relationships between men and women. It is not at all uncommon to hear stories, in media

and in real life, of how men ‘just can’t help themselves’ around beautiful women, and to hear

how devastating it is for men to be rejected by the women to whom they are attracted.”

Although the effect of this code was only a problem because male characters were making

themselves less useful, the cause of this bug had an interesting source in the ideology of the

developer.

By the time of the official release of RimWorld in 2018, this relationship system had changed

to include bisexual male characters, heterosexual female characters, and a more complicated

system of attraction and romancing. Though it is pretty funny, the interesting part of this

225

kerfuffle is not the details of the sexuality coding per se, but the debates it prompted. Because

female characters rarely hit on anyone, and because male characters always hit on female

characters, female characters almost always ended up in relationships with male characters.

Gamers argued about whether the female characters really were bisexual because they were

coded to be equally attracted to female and male characters, or whether they were

heterosexual because they were only ever seen by the player to be heterosexual. This

argument about the essential qualities of RimWorld’s characters and who gets to define them

mirrors debates about authorship in other mediums and discussions about bisexual erasure.

But because games have literal codes by which characters operate, and much about

Rimworld’s characters’ behaviour is situational and emergent, the discussion has to be a little

different.

In the comments of Lo’s article, the developer of the game, Tynan Sylvester, argued that the

female characters of the game were not inherently bisexual or gay: they were not inherently

anything. “People tend to think of game characters as people, but they’re not. They don’t have

internal experiences. They only have outward behaviors, and they are totally defined by those

behaviors, because that’s all the player can see, and the player’s POV is the only one that

matters.” (Sylvester, quoted in D'Anastasio 2016). Sylvester argued that his code was not the

definitive arbiter of a character’s deepest yearnings, the player’s experience of the game was,

and RimWorld’s coding gave an impression of heteronormativity to the player. In Sylvester’s

argument, the player gives the final meaning to the characters’ actions, not the developer, and

definitely not the character.

The comments on Lo’s article held a few positions. Most players noted that no matter the

game characters’ identity, the bug was really annoying and had to be fixed (CartonofMilk in Lo;

modzero in Lo; Orillion in Lo; Swanny in Lo). They also conceded that the game was still in

development and still had a lot of work to be done on it, so the current flaws should be seen as

temporary glitches (cocoarico in Lo; meeper in Lo; Revil in Lo; Swanny in Lo; Walsh in Lo).

Many argued that fictional worlds, and especially computer games, do not have to reflect

reality in their social dynamics (aepervius in Lo; Chaotic Entropy in Lo; klops in Lo; Tacroy in

Lo), and furthermore, discussion of fictional characters’ fictional sexuality was really silly

(Calculon in Lo; pepperfez in Lo). Some said that games should be judged only on the merits of

226

their entertainment value rather than their realism (PancakeWizard in Lo); some said games

that model social systems have a responsibility to model them well (flashlight_eyes in Lo;

shde2e in Lo; syndrome in Lo). A few suggested that most games don’t even try to have a

social system, and that the headaches of setting one up and exposing it to critique puts

developers off trying (Cmaster in Lo; Daemoroth in Lo; nickylee in Lo). An argument

developed between people who said that it was important for games to represent women and

queer people accurately (Armillery in Lo; bouchacha in Lo; DoubleG in Lo; Michael Anson in

Lo; Vigil in Lo), people who said it wasn’t (dauw in Lo; NickAragua in Lo), and people who said

that RimWorld’s modelling seemed pretty spot on to them (aepervius in Lo; Faxanadu in Lo;

MrUnimport in Lo; Scelous in Lo). Sylvester said that the coding of his characters reflected his

experience of other people’s bisexuality, in that women were universally bi-curious and

bisexual men were actually gay, and cited a number of surveys that apparently supported his

observation (TynanSylvester in Lo). A widely-shared concern was the effect of various

ideologies on the development of games and representation of characters (bouchacha in Lo;

Lambert2191 in Lo; modzero in Lo). Many people unhappily acknowledged that ideas about

how the real world works and how the real world should work has an effect on the

development and the player experience of fictional worlds.

One of the things at stake in this case is the search for something persistent at the core of

computer game characters. Who gets to be in the search party is a part of that problem and

what characteristics count as an identity is another. Interestingly, Sylvester’s argument that

RimWorld’s female characters presented as heterosexual to players and that’s all that

mattered, did not gain much traction among commenters. The code, for the commenters, held

the truth about the characters’ identities, not the players’ observations. Part of this is the

difficulties of bisexual representation. It’s hard to tell if someone’s bisexual by just looking at

them – you have to ask them. NPCs cannot be asked. They are always defined by someone else,

and their basic definition is not a player character. As beings defined by a dualistic negation,

it’s not too surprising that nonplayer characters aren’t fully realised, well-rounded individuals.

They are representatives of a system rather than individuals, embodiments of game

mechanics, and so the enjoyment for the player comes from figuring out and manipulating the

system rather than deeply appreciating its components. The true identity of NPCs is slippery,

unable to be pinpointed, and so inherently queer in its frustrating (for some) or unimportant

227

(for others) absence. “The problem,” as Brice succinctly puts it, “is getting people to care about

the subject on top of which a system is mapped” (78), but game characters can never escape

the kind of power that Foucault calls subjectification – the assigning and fixing of identity.

They are always going to be defined by what they appear to be to others, whether by game

code or appearance.

RimWorld’s genre has to be taken into account when discussing the fundamental inner truths

of its characters. RimWorld is a simulation game, and has no linear storyline: once the initial

events are set up it’s up to the player to stop/help the world falling into chaos in their own

preferred way. The player is a god-like figure who has no avatar in the game world, but can

assign tasks to and view statistics on individual astronauts. Simulations like RimWorld do not

have an established series of events but rely on players to construct their own personal

narratives from the events that emerge from the gameplay. Sylvester’s call to define characters

from their behaviour over time rather than from their predetermined attributes follows from

his creation of a story that cannot be described beforehand, only experienced through.

Because RimWorld’s characters are affected by random events and change over time,

Sylvester’s resistance to the idea that they have immutable identities is understandable. The

game and its characters are supposed to be about the player defining their world for

themselves. However, the predetermined attributes Sylvester did program resulted in a world

full of characters who unproductively hit on lesbians all day. As representations go, it isn’t a

totally inaccurate one. Orkin posits that “today, players expect more realism, to complement

the realism of the physics and lighting in the environments” (3), but game characters only

present a drag realness. Their identity is made in concert with those that observe them,

individuate them and subjectify them.

The problems of nonplayer characters comes in part from trying to represent them as human-

like but still designing them to be essentially inferior to the player. “Animals and robots are

generally seen as lacking in some respect by comparison to humans; they have less value and

less or no moral status.” Johnson and Verdicchio remark (292). Coupled with Paepcke ‘s

observation that “the expectations we bring to interacting with robots will influence our final

perceptions of those robots and the ways in which we interact with them.” (4), it seems like

nonplayer characters are defined in relation to humans before they even exist. Despite

228

advances in AI and animation, the emotions of nonplayer characters just aren't as affective to

players as the emotions of another human. This isn't just a case of graphics quality or snappy

dialogue: while players can accurately recognise the kinds of emotions nonplayer characters

perform, they also recognise that nonplayer characters are machines programmed to fulfil a

role, rather than beings with genuine desires and inner worlds. But a player being affected by

nonplayer characters exactly as they would be by a fellow human is not a realistic standard.

Nonplayer characters affect players in unique ways, and this unique affection is wonderful in

itself. Opportunities to relate to NPCs in new ways can be found in their repetitiveness, their

reliability, and their strange senses, without essentialising them.

THEY AFFECT PEOPLE

NPCs do affect humans, though perhaps not in the way everyone would like. Importantly for

this thesis, humans affect NPCs too. A nonplayer character does not have human emotions

because it does not have human awareness of the ways in which it is being affected. But

nonplayer characters do register some affect, in some limited and unfamiliar ways. When a

game starts or an area loads, some parts of an NPC are activated, waiting for the player to

come across them. When an NPC is touched, its hitbox registers a collision, information that is

given or taken by other processes. When an NPC is within the camera’s view, it acquires a

rendered body and an increased ability to act. When an NPC is triggered into attacking or

chatting, it follows a branching series of actions. When an NPC dies, it is instantly gone, or

instantly replaced with an inert representation of itself.

These are not isolated processes but done in concert with the player. The NPC is affected by

the player’s actions and the player is affected by the NPC’s reactions, as well as their

anticipation and memory of similar events. Despite what Spinoza says about adequate and

inadequate ideas, actions and reactions are not so easy to tease apart, even for the most self-

aware. von UexkuH ll helps in this regard through his theory of the umwelt, or how inner worlds,

or more specifically the subjective perceptions of affordances, work. Just as nonhuman

229