Top Banner
1 CIG 09 Milan, Sept 2009 A Task Annotation Model for SandBox Serious Games F. Bellotti, R. Berta, A. De Gloria, L. Primavera ELIOS Lab DIBE – University of Genoa Elios Lab - UniGe
29
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Page 1: CIG 2009

1CIG 09

Milan, Sept 2009

Click to edit Master title style

A Task Annotation Model for SandBox Serious Games

F. Bellotti, R. Berta, A. De Gloria, L. Primavera

ELIOS Lab DIBE – University of Genoa

Elios Lab - UniGe

Page 2: CIG 2009

2CIG 09

Milan, Sept 2009

Click to edit Master title style

SummaryA specialization of computer gamesInfuse instruction into the game play experience

Huge potentiala growing slice of population is familiar with playing gamesRealistic and compelling challenges, capturing user concentration span

for long timessuited to engage the learner and encourage active construction of

meaning and development of skills

Methodologies/tools now needed to support pedagogical authors in developing effective learning experiences

Serious Games

Page 3: CIG 2009

3CIG 09

Milan, Sept 2009

Click to edit Master title style

Elios and partnerImportance of helping students develop well-connected knowledge structures representing relationships among facts and concepts

When the knowledge structure for a topic is well-connected, new information is more readily acquired, since the cognitive load is low

The Sand Box Serious Games (SBSGs) SG class

Successful pure entertainment gamesGrand Theft Auto, Oblivion

Cognitive Processing

Page 4: CIG 2009

4CIG 09

Milan, Sept 2009

Click to edit Master title styleThe SBSG model

A concrete spatial organization where knowledge is distributed

The VW to be explored

Tasks, spread and contextualized in the VWTasks embody units of knowledge that can be discovered and accessed by the player in order to construct meaning, build lasting memories, deepen understanding

Task Based Learning (TBL) theoryConcrete, focused activities to construct knowledge and develop

skills Characterizing tasks:

– An ability to engage the learner’s interest; a primary focus on meaning; a need to be completed; an outcome in terms of which the success is judged; a clear relationship with real-world activities

Page 5: CIG 2009

5CIG 09

Milan, Sept 2009

Click to edit Master title styleInserting tasks in SBSGs

Content

Delivery StrategyWhen, where and how they become are assigned to the player

A key aspect to support effective knowledge buildingUsually tackled through author-scripted levels and missions

– Costly and difficult to scale

Page 6: CIG 2009

6CIG 09

Milan, Sept 2009

Click to edit Master title styleVision

Hp: huge availability of simple contents (tasks)Elementary online visual authoring toolsAlso User Generated Contents (UGC) trend.

Simple task models:Minigames, quizzes, visiting a limited VW zone, interactions with

3d objects or CVHs

Tasks are semantically annotated by the authorEmployable in various VW environmentsA city, an industrial plant, an island

Page 7: CIG 2009

7CIG 09

Milan, Sept 2009

Click to edit Master title styleDecoupling contents from their delivery strategy

Domain experts focus on defining the tasks’ content and annotation

e.g. topics, difficulty, geo-ref. position, learning style, etc.

Game authors specify the requirements for a runtime task delivery strategy according to their educational and entertainment objectives

Strategy learned by a specific CI module embedded in a Game Engine (GE)

This Experience Engine (EE) performs the gameplay task sequencingdrive the player’s exploration of the environment by dynamically

assigning tasks tailored on his actual profile

Page 8: CIG 2009

8CIG 09

Milan, Sept 2009

Click to edit Master title styleWorkflow and high-level System Architecture

Page 9: CIG 2009

9CIG 09

Milan, Sept 2009

Click to edit Master title styleConcept (I)

No need for a detailed story specification

A high-level challenge spurring competition through exploration

E.g. various types of treasure hunts

The story emerges from player-environment, player-task, player-player interactions in a meaningful (3d) context

A shift of focus is needed, in content production, from narrative to actual edu contents (the tasks)

Enlarge the basis of authors and provide users with high quality, content-rich multimedia educational experiences.

Page 10: CIG 2009

10CIG 09

Milan, Sept 2009

Click to edit Master title styleConcept (II)

Not better than a strong interactive narrative architecture

They are complementary approaches

But it is a new model that uses AI to deal with a database of semantically annotated tasks and with various types of user features/needs and teacher perspectives/requirements

VW independent

Page 11: CIG 2009

11CIG 09

Milan, Sept 2009

Click to edit Master title styleThe model (I)

SBSG as an exploration field where the player is given missions

Missions consists of tasks embedded in the VW

Page 12: CIG 2009

12CIG 09

Milan, Sept 2009

Click to edit Master title styleThe model (II)

The requirements for the decision policy are parameters of the model

Values to be specified by the game author

The runtime system will learn the decision policy

Choice of the tasks according to their fitness to the author strategy with regard to the player

Page 13: CIG 2009

13CIG 09

Milan, Sept 2009

Click to edit Master title styleAnnotating the tasks

Proposed parametersType (quiz, arcade, dialogue, exploration)

Entertainment value

Quality value

Skill relevance

Skill benefits

Difficulty

Covered learning styles

Dependences

Place

Page 14: CIG 2009

14CIG 09

Milan, Sept 2009

Click to edit Master title styleSample values

T1: Wind-Rose Puzzle

T2: Nodes Quiz Easy

T3: Weak-Wind Test

T4: Strong-Wind Test

Task Type Puzzle 100%

Quiz 80% VisualQuiz 20%

Arcade 80% Quiz 20%

Arcade 100%

Entertainm. 5 7 7 10 Skill Relevance

Boats 50% Winds 50%

Boats 100%

Boats 70% Winds 30%

Boats 80% Winds 20%

Covered Learning Styles

Visual 100%

Textual 80% Visual 20%

Simulation 80% Text 20%

Simulation 100%

Difficulty 5 4.45 8 8.5 Difficulty Adapt. Range

± 1 ± 3 ± 2 ± 2

Page 15: CIG 2009

15CIG 09

Milan, Sept 2009

Click to edit Master title styleUser Model

Skill Level

Navigation Ability Level

Task Type Need/Preference

Skill Need/Preference

Learning Style Need/Preference

Visual, auditory, reading and kinesthetic learning modalities

Page 16: CIG 2009

16CIG 09

Milan, Sept 2009

Click to edit Master title styleModel of the Delivery Strategy

Definition of a Cost Function for a Mission

To be minimized by the runtime system

Cost = deviation from the reference values specified by the game author

Page 17: CIG 2009

17CIG 09

Milan, Sept 2009

Click to edit Master title styleReference Difficulty Curve

Page 18: CIG 2009

18CIG 09

Milan, Sept 2009

Click to edit Master title style

Navigation Difficulty Curve

Inter-task dependences

CNoDep penalizes the absence of dependences from previous tasks (Nda 0 or 1)

CFailedDep penalizes the absence, in the previous tasks of the sequence, of a task on which the ath task depends (Fda 0 or 1)

CIs

Page 19: CIG 2009

19CIG 09

Milan, Sept 2009

Click to edit Master title styleDistributions

Task Type

Learning Style

Covered Skills

Page 20: CIG 2009

20CIG 09

Milan, Sept 2009

Click to edit Master title styleCIs

Entertainment and quality thresholds

Trend for entertainment and quality

Performance curve

Page 21: CIG 2009

21CIG 09

Milan, Sept 2009

Click to edit Master title styleSkillBenefit Threshold

The benefit for the player must be above a threshold

Repetition

Page 22: CIG 2009

22CIG 09

Milan, Sept 2009

Click to edit Master title styleRuntime AI Engine

EE built atop of the Torque Game EngineGenetic Algorithm implementationThe EE can operate also in absence of the specification of the delivery strategy

Continuous adaptation of the game flow without aiming at the achievement of target knowledge levels predefined by the authorReinforcement Learning implementation

Page 23: CIG 2009

23CIG 09

Milan, Sept 2009

Click to edit Master title styleSample Application: Travel in Europe

The TiE platform is designed to promote the knowledge of the European cultural heritage to a wide audience using the paradigm of multiplayer SGsA cultural treasure hunt across Europe

Visit a certain number of cities (objective), and in each town the player has one or more missions to accomplish.

Page 24: CIG 2009

24CIG 09

Milan, Sept 2009

Click to edit Master title styleSample Application: Travel in Europe

Mission are characterized by a number of questionsThe player’s explores the urban environments in search of the places indicated in the missionAt each target PoI, the player looks for 3D iconsEach icon triggers a task session through which the user can virtually manipulate pieces of the artistic heritage and face quizzes related to the PoI

Page 25: CIG 2009

25CIG 09

Milan, Sept 2009

Click to edit Master title styleSample Application: Travel in Europe

The structure and high-level plot is quite simple and schematic

But it is a scalable and flexible formatCost-effective development (3rd parties & UGC)

Little overhead on the player, who can focus his attention on the contents

Easy to understand overall structure

Minigames instances of templates

3D settings, game mechanics and interaction modalities necessarily high quality

Page 26: CIG 2009

26CIG 09

Milan, Sept 2009

Click to edit Master title styleConclusions (I)

Methodologies to support effective authoring of SGs

Abstract modelVW , tasks

Support authors to focus on defining the tasks

No game scriptinghard-coding events, actions, locations and contents of

a specific game.

Page 27: CIG 2009

27CIG 09

Milan, Sept 2009

Click to edit Master title styleConclusions (II)

Task annotation allows reusing tasks also in different games

UGC: involvement of user communities

A new authoring modelExploit repositories of tasksBuild educational games specifying a VW and the task

delivery strategy criteria for the missions– Learnt at runtime by a CI engine.

New perspectives to pedagogical experts and game designers to provide players with adaptive contents

Page 28: CIG 2009

28CIG 09

Milan, Sept 2009

Click to edit Master title styleUpcoming work

Extensive user testing is now neededAssess and analyze the support for effective learning in a real entertainment context

Open issuesAn extensive validation of the user and task models and of the delivery strategy cost function

The speed of the runtime user adaptation process

However, it is important that the developed games are able to attract users, similarly to commercial videogames, even if they deal with educational topics

Page 29: CIG 2009

29CIG 09

Milan, Sept 2009

Click to edit Master title styleThank you!