NATURALISM VS. STYLIZATION IN STATIC VISUAL APPEARANCE + Results from PILOT STUDY ( June/July-06 ) VIRTUAL PEDAGOGICAL AGENTS ( VPAs ) Static Visual Appearance Static visual appearance are the inanimate qualities of a virtual pedagogical agent (VPA), brought to life by visual dynamics to make up a total visual experience. – Static Qualitities: graphical style, body, face, clothes, attributes, colors, textures – Dynamic Qualities: movements, gestures, facial expressions, gaze ► Read more in [ 4 ] DEGREE OF HUMANNESS BASIC PHYSICAL PROPERTIES (E.g. Body Types) GRAPHICAL STYLE Naturalism – Stylization GRAPHICAL STYLE Detailed – Simplified Figure 1. Dimensions of Static Visual Appearance Naturalism vs. Stylization Naturalism vs. Stylization: Dimensions Graphical style is one of the dimensions of static visual appearance. Two graphical style aspects are of particular interest in virtual agents. Naturalism vs. Stylization: e scope of stylization spans over a wide range of styl- ized expressions (see Figure 1). Detailedness vs. Simplification: Consider the reduction of a photo to different levels of details: – It could be turned into a semi-detailed (semi-naturalistic) contour drawing. – It could be transformed into a very simplified (and stylized) cartoon. ese example illustrate that there are no simple linear relations in the design space of naturalism – stylization. ► Read more in [ 4 ] Figure 2. The pictorial space of realism, iconicity and abstraction Naturalism vs. Stylization: Arguments In favour of visual naturalism in VPAs the literature presents three arguments: (i) the smooth communication argument (ii) the identification argument (iii) the believability argument In favour of visual stylization in VPAs there are: (i) the false expectation argument (ii) the problem of representing perfection argument When scrutinized, however, none of the arguments hold. Furthermore, empirical data is sparse, and some studies are too imprecise in their handling of visual details and variables to actually have any bearing at all on the complex visual spaces of naturalism vs. stylization. ► Read more in [ 4 ] Naturalism vs. Stylization: Motives Behind conflicting arguments on visual naturalism versus stylization we find differ- ent motives for research and development of VPAs. – One is to obtain smooth and beneficial interaction between humans and comput- ers in pedagogical contexts – Another is to model and understand human behavior (human dialogues, gestures, facial expressions) With respect to the second motive but not the first, naturalism is taken to be the self-evident ultimate goal. When researching naturalism vs. stylization in VPAs a distinction should be made between the scientific modelling of naturalistic human behavior and the pragmatic approach focusing on the development of a usable tool. Only this way can we benefit fully, both from the potentials of VPAs as powerful test-beds for theoretical modelling and for their pedagogical potentials in role mod- elling, identification, and stereotype use. The two dimensions above of Graphical Style: ‘Naturalism - Stylization’ & ‘Detaildness - Simplification’ in Static Visual Appear- ance (see Figure 1) can be related to a pictorial space as described by Scott McCloud in his book Understanding Comics: The Invisible Art [ 7 ] Pilot Study Pilot Study: Design Study procedure – Twenty 13-19-year-old participants (9 female, 11 male) – An educational chat environment – Choice of an avatar out of eight paired characters, four stylized and four natural- istic (see Figure 3 below) – An interview aimed at illuminating the influences of avatar choice with respect to naturalism vs. stylisation, in particular similarity with oneself in different senses, i.e. issues with bearing on the identification argument in favour of visual realism mentioned above Study design guidelines e agents – e user group being teens, the choice fell on two established and successful graphical styles within this group: e styles of Manga and of the Sims – Focusing on naturalism vs. stylization, other variables known to impact an agent/avatar choice were held relatively constant among all the avatars Avatar choice and reasoning about avatars – Embedded in a relatively rich and natural application context involving other choices, actions and questions Figure 3. Screen shot: page for ‘Chat Agent’ selection Pilot Study: Results Choice of avatar – Stylized (Manga style): 8 – Naturalistic (Sims style): 12 e distribution was even for both age and gender. Yet both of these parameters will be of interest to follow in the main study, since previous studies show signifi- cant gender differences: females seem more likely than males to chose stylized over naturalistic characters. [1,2,3] Motives for choice – About 1/3 of the participants emphasize similarity to themselves by personality or by appearance (the latter only for naturalistic characters) – 1/4 emphasize appearance and/or personality without reference to similarity with themselves – About 1/3 of the participants, of which most chose a stylized character, seem to use the possibility to “look different” or “be different” by choosing a character they place among the four least similar to themselves, or least ‘like me’: - one chose a naturalistic character (“I chose this bloke who looks really good”) - five chose stylized characters (e.g. “she looks open and extrovert, not like me”; “it would be nice to have a little hair-cut like that, I surely don’t look like that now, but I’d like to”) Discussion: It is possible that stylized characters provide more opportunities for being someone else and someone desired, since it is easier to read yourself into a stylized character (cf. theories by McCloud [ 7 ] ) which leaves more to imagination than a naturalistic character (Figure 4). Figure 4. Stylized characters may be easier to adopt and identify with than naturalistic ones Comparison of avatar styles Late on during the interview subjects were explicitly asked to compare the two dif- ferent styles of characters as to positive and negative features. – As a positive feature of the naturalistic characters, 6 participants emphasize their choice of realism as such (e.g. “I believe these are better as they look more natu- ral”; “they are realistic”; “they look more real”; “I surely prefer one that looks like me over a comics figure”) – 3 participants underline that the faces of the naturalistic characters are more dis- tinct – 2 participants hold that the naturalistic characters look better and are better de- signed (e.g. “these are well done”; “these are more attractive simply”) – e non-realism is, on the other hand, held forth as a positive feature of the stylized characters by 4 participants (e.g. “there is more for your own imagination here, and with these you can express your personality instead, what you have inside”; “these really concern personality”; “with these, the Manga ones, you can be who- ever you like and you don’t have to reveal yourself to people that you don’t know”) – 3 participants turn the ’looking good’ argument around (e.g. “your face should not be important, and to be too good-looking”; “I prefer these , because those [the naturalistic] seem more obsessed with their look… and me … well I am a bit obsessed, but not so much”) – 6 participants emphasize the stylized characters as being more ‘fun’ (e.g. “one ad- vantage with these is that it doesn’t feel as rigid”; “these are a bit more fun as characters”, “these are more fun, those [naturalistic] are serious in a strange way, I see no advantage with that”; “these are more fun and are prettier, it’s so cool with such big eyes”) – 2 participants to the contrary emphasize the advantage of the naturalistic charac- ters as being more serious (e.g. “it is an advantage with these that you feel more serious about what you are doing”; “the others are too much like comics”) Issue for follow-up An issue to be pursued in the main study is the possible relations between: a) participants choice of graphical style (naturalistic vs. stylized characters) b) participants social/communicative style (task oriented versus relation orien- tated) e results from the pilot study are not statistically significant but the trends line up with some of our previous results. [ 3 ] Figure 5. Social/communicative style vs. graphical style Pilot Study: Summary Overall diversity A central result of the pilot study was diversity in choices: of avatars, and in motives for the choices and opinions on the two kinds of graphical styles – a diversity spreading over gender and age. A striking result was also that the majority of the participants easily found advan- tages and disadvantages with both styles. Furthermore, the 8 last participants were asked for their view on the alternative ‘keeping to one style’ or ‘mixing styles’ of avatars, as in the prototype they had tested. All eight went for ‘mixing’, most of them with conviction in tone and content of their answer. Pilot Study: Comments – e pilot study concerns avatars in a pedagogical chat context. It does not ad- dress pedagogical agents in the roles of virtual teachers, coaches, learning com- panions, etc. e role of an agent seems to influence users preferences as to visu- ally naturalistic or visually stylized characters. [5] – e study does not address interactivity and user responses towards visually more or less naturalistic characters that behave more or less naturalistically. – Similarity in appearance seems to be a factor that influences the choice of an avatar for some participants but not others. Analyzing the answers as to in what way a character looks similar to the person, the hair (haircut, hairstyle, color of hair) took a clear lead; thereafter eyes and face were mentioned. e clothing was not referred to by any participant, the reason probably being that the clothing is held relatively constant among half-length characters. Compare this outcome to the study by Hall et al. [6], where clothing was a main parameter in evaluating the role of appearance similarity for emotional responses in young users. 0 1 2 3 4 5 Naturalistic Character Stylized Character Uncertain Relation Oriented Task Oriented References [1] Baylor, A. (2005). e impact of pedagogical agent image on affective outcomes. In Proc. of Workshop on Affective Interac- tions: Computers in the Affective Loop, Int. Conf. on Intelligent User Interfaces (IUI’05), San Diego, CA. [ 2 ] Gulz, A. (2005). Social enrichment by virtual characters – differential benefits. J. of Computer Assisted Learning, 21, (6), 405-418. [3] Gulz, A. & Haake, M. (2005). Social and visual style in vir- tual pedagogical agents. In Proc. of Workshop on Adapting the Interaction Style to Affective Factors, Int. Conf. on User Model- ing (UM’05), Edinburgh, Scotland. [ 4 ] Gulz, A. & Haake, M. (2006). Visual design of virtual peda- gogical agents: Naturalism versus stylization in static appear- ance. Paper to be presented at the Int. Design and Engagabil- ity Conference @ NordiCHI 2006 (iDec3), Oslo, Norway. [ 5 ] Haake, M. & Gulz, A. (2005). A look at the roles of look & roles in virtual pedagogical characters. Paper that was to be presented at the Int. Conf. on Intelligent User Interfaces (IUI’06), Sidney, Australia. [ 6 ] Hall, L., Woods, S., Dautenhahn, K., Sobral, D., Paiva, A., Wolke, D. & Newall, L. (2004). Designing empathic agents: Adults versus kids. In Proc. of the Int. Conf. on Intelligent Tu- toring Systems (ITS’04), Maceió, Alagoas, Brazil, Springer- Verlag, Germany, 604-613. [7] McCloud, S. (1993). Understanding Comics: e Invisible Art. HarperPerennial, New York, NY. Footnotes None of these being 2D vs. 3D. ► Read more in [ 4 ] ey were used as an inspiration. No more than 3-4 partici- pants identified this Sims and/or Manga inspiration, even though most of them (according to the interview) seemed fa- miliar with Manga and the Sims. Two male participants, aged 17 and 19, chose a naturalistic avatar, but were explicit in that stylized characters were their overall preference. One spoke at length about their advan- tages. Another indicated that he might have remade his choice, if possible, into a stylized character. Yet both com- mented early on in the interview that the stylized characters looked young or childish in comparison to the others. is in- dicates that we did not succeed completely in keeping the age factor as constant as aimed for. (However, some participants aged 16-19 did choose a stylized character.). In the pilot study social/communicative style was measured with a version of Galotti et al’s. questionnaire ATTLS. Only participants with a previously set difference, in any direction between task and relation orientation were included. • Agneta Gulz, Associate professor Div. of Cognitive Science (LUCS) Lund University [email protected] • Magnus Haake, PhD student Dept. of Design Sciences, LTH Lund University [email protected] • Daniel Labbé, Research assistant Div. of Cognitive Science (LUCS) Lund University [email protected]
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
NATURALISM VS. STYLIZATION IN STATIC VISUAL APPEARANCE + Results from PILOT STUDY (June/July-06)V
IRT
UA
L P
ED
AG
OG
ICA
L A
GE
NT
S (
VP
As) Static Visual Appearance
Static visual appearance are the inanimate qualities of a virtual pedagogical agent (VPA), brought to life by visual dynamics to make up a total visual experience.
– Static Qualitities: graphical style, body, face, clothes, attributes, colors, textures
– Dynamic Qualities: movements, gestures, facial expressions, gaze
► Read more in [4]
DEG
REE O
FH
UM
AN
NESS
BA
SIC PH
YSIC
AL
PRO
PERTIES
(E.g. B
od
y Types)
GR
APH
ICA
L STY
LEN
aturalism
– Stylization
GR
APH
ICA
L STY
LED
etailed – Sim
plifi
ed
Figure 1. Dimensions of Static Visual Appearance
Naturalism vs. Stylization
Naturalism vs. Stylization: DimensionsGraphical style is one of the dimensions of static visual appearance. Two graphical style aspects are of particular interest in virtual agents.
Naturalism vs. Stylization: The scope of stylization spans over a wide range of styl-ized expressions (see Figure 1).
Detailedness vs. Simplification: Consider the reduction of a photo to different levels of details:– It could be turned into a semi-detailed (semi-naturalistic) contour drawing.– It could be transformed into a very simplified (and stylized) cartoon.
These example illustrate that there are no simple linear relations in the design space of naturalism – stylization.
► Read more in [4]
Figure 2. The pictorial space of realism, iconicity and abstraction
Naturalism vs. Stylization: ArgumentsIn favour of visual naturalism in VPAs the literature presents three arguments:(i) the smooth communication argument(ii) the identification argument(iii) the believability argument
In favour of visual stylization in VPAs there are:(i) the false expectation argument(ii) the problem of representing perfection argument
When scrutinized, however, none of the arguments hold. Furthermore, empirical data is sparse, and some studies are too imprecise in their handling of visual details and variables to actually have any bearing at all on the complex visual spaces of naturalism vs. stylization.
► Read more in [4]
Naturalism vs. Stylization: MotivesBehind conflicting arguments on visual naturalism versus stylization we find differ-ent motives for research and development of VPAs.– One is to obtain smooth and beneficial interaction between humans and comput-
ers in pedagogical contexts– Another is to model and understand human behavior (human dialogues, gestures,
facial expressions)
With respect to the second motive but not the first, naturalism is taken to be the self-evident ultimate goal. When researching naturalism vs. stylization in VPAs a distinction should be made between the scientific modelling of naturalistic human behavior and the pragmatic approach focusing on the development of a usable tool. Only this way can we benefit fully, both from the potentials of VPAs as powerful test-beds for theoretical modelling and for their pedagogical potentials in role mod-elling, identification, and stereotype use.
The two dimensions above of Graphical Style: ‘Naturalism − Stylization’ & ‘Detaildness −
Simplification’ in Static Visual Appear-ance (see Figure 1) can be related to
a pictorial space as described by Scott McCloud in his book
Understanding Comics: The Invisible Art [7]
Pilot Study
Pilot Study: Design
Study procedure– Twenty 13-19-year-old participants (9 female, 11 male)– An educational chat environment– Choice of an avatar out of eight paired characters, four stylized and four natural-
istic (see Figure 3 below)– An interview aimed at illuminating the influences of avatar choice with respect
to naturalism vs. stylisation, in particular similarity with oneself in different senses, i.e. issues with bearing on the identification argument in favour of visual realism mentioned above
Study design guidelines
The agents– The user group being teens, the choice fell on two established and successful
graphical styles within this group: The styles of Manga and of the Sims – Focusing on naturalism vs. stylization, other variables known to impact an
agent/avatar choice were held relatively constant among all the avatars
Avatar choice and reasoning about avatars – Embedded in a relatively rich and natural application context involving other
choices, actions and questions
Figure 3. Screen shot: page for ‘Chat Agent’ selection
Pilot Study: Results
Choice of avatar– Stylized (Manga style): 8– Naturalistic (Sims style): 12The distribution was even for both age and gender. Yet both of these parameters will be of interest to follow in the main study, since previous studies show signifi-cant gender differences: females seem more likely than males to chose stylized over naturalistic characters. [1,2,3]
Motives for choice– About 1/3 of the participants emphasize similarity to themselves by personality or
by appearance (the latter only for naturalistic characters)– 1/4 emphasize appearance and/or personality without reference to similarity
with themselves– About 1/3 of the participants, of which most chose a stylized character, seem to
use the possibility to “look different” or “be different” by choosing a character they place among the four least similar to themselves, or least ‘like me’:- one chose a naturalistic character (“I chose this bloke who looks really good”)- five chose stylized characters (e.g. “she looks open and extrovert, not like me”;
“it would be nice to have a little hair-cut like that, I surely don’t look like that now, but I’d like to”)
Discussion: It is possible that stylized characters provide more opportunities for being someone else and someone desired, since it is easier to read yourself into a stylized character (cf. theories by McCloud [7]) which leaves more to imagination than a naturalistic character (Figure 4).
Figure 4. Stylized characters may be easier to adopt and identify with than naturalistic ones
Comparison of avatar stylesLate on during the interview subjects were explicitly asked to compare the two dif-ferent styles of characters as to positive and negative features.– As a positive feature of the naturalistic characters, 6 participants emphasize their
choice of realism as such (e.g. “I believe these are better as they look more natu-ral”; “they are realistic”; “they look more real”; “I surely prefer one that looks like me over a comics figure”)
– 3 participants underline that the faces of the naturalistic characters are more dis-tinct
– 2 participants hold that the naturalistic characters look better and are better de-signed (e.g. “these are well done”; “these are more attractive simply”)
– The non-realism is, on the other hand, held forth as a positive feature of the stylized characters by 4 participants (e.g. “there is more for your own imagination here, and with these you can express your personality instead, what you have inside”; “these really concern personality”; “with these, the Manga ones, you can be who-ever you like and you don’t have to reveal yourself to people that you don’t know”)
– 3 participants turn the ’looking good’ argument around (e.g. “your face should not be important, and to be too good-looking”; “I prefer these , because those [the naturalistic] seem more obsessed with their look… and me … well I am a bit obsessed, but not so much”)
– 6 participants emphasize the stylized characters as being more ‘fun’ (e.g. “one ad-vantage with these is that it doesn’t feel as rigid”; “these are a bit more fun as characters”, “these are more fun, those [naturalistic] are serious in a strange way, I see no advantage with that”; “these are more fun and are prettier, it’s so cool with such big eyes”)
– 2 participants to the contrary emphasize the advantage of the naturalistic charac-ters as being more serious (e.g. “it is an advantage with these that you feel more serious about what you are doing”; “the others are too much like comics”)
Issue for follow-upAn issue to be pursued in the main study is the possible relations between:a) participants choice of graphical style (naturalistic vs. stylized characters)b) participants social/communicative style (task oriented versus relation orien-
tated)The results from the pilot study are not statistically significant but the trends line up with some of our previous results. [3]
Figure 5. Social/communicative style vs. graphical style
Pilot Study: Summary
Overall diversityA central result of the pilot study was diversity in choices: of avatars, and in motives for the choices and opinions on the two kinds of graphical styles – a diversity spreading over gender and age.
A striking result was also that the majority of the participants easily found advan-tages and disadvantages with both styles. Furthermore, the 8 last participants were asked for their view on the alternative ‘keeping to one style’ or ‘mixing styles’ of avatars, as in the prototype they had tested. All eight went for ‘mixing’, most of them with conviction in tone and content of their answer.
Pilot Study: Comments
– The pilot study concerns avatars in a pedagogical chat context. It does not ad-dress pedagogical agents in the roles of virtual teachers, coaches, learning com-panions, etc. The role of an agent seems to influence users preferences as to visu-ally naturalistic or visually stylized characters. [5]
– The study does not address interactivity and user responses towards visually more or less naturalistic characters that behave more or less naturalistically.
– Similarity in appearance seems to be a factor that influences the choice of an avatar for some participants but not others. Analyzing the answers as to in what way a character looks similar to the person, the hair (haircut, hairstyle, color of hair) took a clear lead; thereafter eyes and face were mentioned. The clothing was not referred to by any participant, the reason probably being that the clothing is held relatively constant among half-length characters. Compare this outcome to the study by Hall et al. [6], where clothing was a main parameter in evaluating the role of appearance similarity for emotional responses in young users.
0
1
2
3
4
5
Naturalistic Character
Stylized Character
UncertainRelationOriented
TaskOriented
References
[1] Baylor, A. (2005). The impact of pedagogical agent image on affective outcomes. In Proc. of Workshop on Affective Interac-tions: Computers in the Affective Loop, Int. Conf. on Intelligent User Interfaces (IUI’05), San Diego, CA.
[2] Gulz, A. (2005). Social enrichment by virtual characters – differential benefits. J. of Computer Assisted Learning, 21, (6), 405-418.
[3] Gulz, A. & Haake, M. (2005). Social and visual style in vir-tual pedagogical agents. In Proc. of Workshop on Adapting the Interaction Style to Affective Factors, Int. Conf. on User Model-ing (UM’05), Edinburgh, Scotland.
[4] Gulz, A. & Haake, M. (2006). Visual design of virtual peda-gogical agents: Naturalism versus stylization in static appear-ance. Paper to be presented at the Int. Design and Engagabil-ity Conference @ NordiCHI 2006 (iDec3), Oslo, Norway.
[5] Haake, M. & Gulz, A. (2005). A look at the roles of look & roles in virtual pedagogical characters. Paper that was to be presented at the Int. Conf. on Intelligent User Interfaces (IUI’06), Sidney, Australia.
[6] Hall, L., Woods, S., Dautenhahn, K., Sobral, D., Paiva, A., Wolke, D. & Newall, L. (2004). Designing empathic agents: Adults versus kids. In Proc. of the Int. Conf. on Intelligent Tu-toring Systems (ITS’04), Maceió, Alagoas, Brazil, Springer-Verlag, Germany, 604-613.
[7] McCloud, S. (1993). Understanding Comics: The Invisible Art. HarperPerennial, New York, NY.
Footnotes
None of these being 2D vs. 3D.► Read more in [4]
They were used as an inspiration. No more than 3-4 partici-pants identified this Sims and/or Manga inspiration, even though most of them (according to the interview) seemed fa-miliar with Manga and the Sims.
Two male participants, aged 17 and 19, chose a naturalistic avatar, but were explicit in that stylized characters were their overall preference. One spoke at length about their advan-tages. Another indicated that he might have remade his choice, if possible, into a stylized character. Yet both com-mented early on in the interview that the stylized characters looked young or childish in comparison to the others. This in-dicates that we did not succeed completely in keeping the age factor as constant as aimed for. (However, some participants aged 16-19 did choose a stylized character.).
In the pilot study social/communicative style was measured with a version of Galotti et al’s. questionnaire ATTLS. Only participants with a previously set difference, in any direction between task and relation orientation were included.
• Agneta Gulz, Associate professorDiv. of Cognitive Science (LUCS)Lund [email protected]
• Magnus Haake, PhD studentDept. of Design Sciences, LTHLund [email protected]
• Daniel Labbé, Research assistantDiv. of Cognitive Science (LUCS)Lund [email protected]
Related Documents
