Analysis of the Effects of Appearances of Avatars on User's

Self-evaluation of Extroversion

Tomoko Koda and Ryosuke Oguri

Department of Information Science and Technology, Osaka Institute of Technology, Osaka, Japan

Keywords: Agents, Avatars, Computer-mediated Communication, Proteus Effect, Virtual Environments, Serious

Games, Transformed Social Interaction, Avatar Chat, Appearances, Personality.

Abstract: Proteus effect is known as a phenomenon in which the personality and behavior of an individual, within

online virtual worlds, is changed by the appearances of their avatar. This paper reports whether Proteus

effect is applicable to accompanying avatars with whom a user's avatar interacts in a virtual world. The

results indicated that Proteus effect was applicable to those who perceived the avatar was their alter-ego,

while those who did not was not affected by the appearance of the avatar. The results suggest the

importance of considering avatar’s and agent’s appearances and user’s personality in serious games, online

virtual worlds, and avatar-mediated online communications.

1 INTRODUCTION

Interacting with other players via avatars in virtual

worlds and serious games have become our daily

activities. There are wide variety of avatar's

appearances, and users freely select avatars and use

them. However, research in computer-mediated

communication reports the impact of virtual space

and avatar's appearance on users. Thaler et al.

investigated gender difference in preference of users'

own photo-realistic avatars' appearances. Their

results suggested the male participants placed an

importance on their avatar's photo-realistic texture,

while the female participants on their avatars' body

shape (Thaler et al. 2018).

Other research shows the effects of avatar's

appearances and virtual world on user's behaviours.

Pan et al. investigated the effect of avatars in a

shared virtual environment. Their results suggested

that avatar-mediated collaboration led to a better

task performance and higher trust to each other than

the non-avatar condition (Pan and Steed, 2017).

Changing the background image and environmental

sound in the 3D avatar chat space affected the

number of pros and cons in the opinions in user

dialogues (Takeuchi et al., 2016). Yee et al.

suggested the appearance of the avatar used in

online communication influences the user's

behavioural characteristics, especially extroversion,

and named the effect as “Proteus effect".

Specifically, in a laboratory immersive virtual

environment, users who operated an attractive avatar

talked frequently about themselves, interacted with

the other avatar more confidently when their avatar

was taller than the other avatar (Yee and Bailenson,

2007).

Yee's consecutive experiment indicated that

Proteus effect is applicable not only in immersive

virtual spaces but also in online communities in a

virtual world, and this effect has been shown to

persist in face-to-face interaction in real space (Yee

et al., 2009). From these findings, there is a

possibility that when attractive avatars are used in

the virtual world, the users continue to perform

interactions and actions with high extroversion in the

virtual world, thereby affecting their behavioural

characteristics in the real world. Another research

that examined the Proteus effect showed that avatar's

clothing in the virtual space (white / black, KKK,

etc.) affected the attitudes and perceptions of the

users by giving them a priming effect (Pena et al.,

2009).

One of the reasons for user's behavioural change

by changing the appearance of avatar is that the user

sees avatar as their own alter-ego. Tokioka states an

avatar is one in which a user him/herself creates

his/her "alter ego" in a virtual world, and it can be

regarded as the user's own "alter ego" in that it is

created and operated by the user (Tokioka, 2014).

232

Koda, T. and Oguri, R.

Analysis of the Effects of Appearances of Avatars on User’s Self-evaluation of Extroversion.

DOI: 10.5220/0007483502320237

In Proceedings of the 11th International Conference on Agents and Artiﬁcial Intelligence (ICAART 2019), pages 232-237

ISBN: 978-989-758-350-6

However, whether the Proteus effect can be found

for users who do not recognize an avatar as their

alter ego, and whether Proteus effect is applicable

only by their own avatar but also by an

accompanying avatar have not been examined.

Therefore, this research aims to examine the

effect of the appearance of the self-operating avatar

and the appearance of the accompanying avatar used

in the virtual world on the user's self-evaluation of

extroversion. Hypothesis 1 is "Those who regard

their self-operating avatar (“self-avatar”) as their

own alter-ego, if their self-avatar and accompanying

avatar are attractive, their self-evaluation of

extroversion becomes higher compared to when the

avatars are not attractive." Hypothesis 2 is "Those

who do not regard their self-avatar as their alter-ego,

their self-evaluation of extraversion does not change

regardless of the self-avatar's or the accompanying

avatar's appearances." In order to verify the

hypotheses, we conducted an experiment using

avatars with various appearances and compared the

users' self-extroversion assessment by BIGFIVE

personality questionnaire.

2 AVATAR DESIGN

We conducted a pre-experiment in order to select

both male and female avatars' appearances with

various attractiveness. The experimental procedure

is as follows:

1) Ten Japanese university students (8 men and 2

women) participated in the pre-experiment, who did

not participate in the main experiment.

2) Eight male avatars and eight female avatars are

created by using Unity Assets. The participants

evaluated each avatar's attractiveness with a 7-point

Likert scale.

3) Avatars with the two highest evaluation scores

and the two lowest scores in average were selected

for the main experiment. The male and female

avatars selected by the pre-experiment are shown in

Figure 1 and 2, respectively.

3 MAIN EXPERIMENT

3.1 Experiment Overview

The main experiment was conducted by participants

operating their self-avatar in a virtual world. The

self-avatar and an accompanying avatar (set

to the

opposite

gender of the participant) converse each other

Figure 1: Male avatars (attractive and unattractive).

Figure 2: Female avatars (attractive and unattractive).

as they move around the virtual downtown. The

virtual downtown and avatars were developed by

Unity 5.4.1, JavaScript and C #. The conversation

scenario is prescribed. The self-avatar and the

accompanying avatar stop by multiple spots. i.e., a

cafe and an art museum in a first-date situation.

They are invited to answer a questionnaire at

eachstop-by spot. The participants (their avatars) can

get some incentive items to make their first date

successful (i.e., a free ticket for the virtual café or

the museum) by answering the questionnaire.

The questionnaires consist of the BIGFIVE

extroversion questions and dummy items (i.e., what

is your favourite food?) so as to make personality

questions as natural as possible. BIGFIVE is widely

used to measure one's personality traits. BIGFIVE

consists of five factors, but we used 12 questions

regarding extroversion factor in this experiment.

Those questions use adjectives, i.e., sociable,

talkative, cheerful, amiable, outgoing, assertive, and

energetic, with 7-point Likert scale (Gough and

Heilbrum, 1983).

The participants are Japanese university students (17

males, 5 females) from 21 to 23 years old. They

were instructed to evaluate the performance of a

questionnaire platform using avatars in a virtual

world. The true objective of the experiment was told

after the experiment. The conditions of the

experiment are the four conditions shown in Table1,

Analysis of the Effects of Appearances of Avatars on User’s Self-evaluation of Extroversion

233

and the conditions are presented to the participants

at random.

Table 1: Four experimental conditions.

Attractive

self-

avata

-attractive

self-avatar

Attractive

accompan

avata

AT/AT

-attractive

accompan

avata

AT/NAT

3.2 Experiment Procedure

The procedure of the experiment is as follows.

1) Pre-experiment questionnaire: Prior to the

experiment, participants who were participating

in another experiment on another day answered

the BIGFIVE personality questionnaire (pre-

questionnaire) in order to assess the nature of the

participants' personality and not to let the

participants notice the true purpose of the

experiment on the day of the main experiment.

2) Main experiment: The participants selected the

most/least attractive avatars as their self-avatars

from the 4 selections of the same gender, and the

most/least attractive avatars as their

accompanying avatars from the 4 selections of

the opposite gender (4 avatars in total).

3) Participants operate their self-avatar according to

the prescribed dialog scenario. Figure 3 shows

an example greeting dialogue. Figure 4 shows a

dialogue and scenes at a museum.

4) Main questionnaire: They answer a questionnaire

at three designated stop-by spots (see Figure 5).

The questionnaire consists of BIGFIVE

extroversion questions and dummy questions.

5) The participants changed the self-avatar and

accompanying avatar, and repeat 3-4. Conditions

are selected at random.

6) Post-questionnaire: After completing all the

conditions, the participants answer a post-

experiment questionnaire. The question was

"whether each participant regarded the avatars as

his/her alter-ego" with a 7- point Likert scale (1:

not at all – 7: very much).

Figure 3: Example of Avatars' Interaction (greetings).

Figure 4: Example of Avatars' Interaction (at a museum).

Figure 5: Questionnaire (in front of the museum):

Questionnaire consists of BIG FIVE extroversion

questions and dummy questions* (6 point scale)

Questions asked above are: I like talking to others. / I

often go out for some activities. / I am talkative. / I like

traveling.* / I am good at expressing myself. / I am shy. / I

like taking photos.*

ICAART 2019 - 11th International Conference on Agents and Artiﬁcial Intelligence

234

4 RESULTS AND DISCUSSION

4.1 Analyses of the Results

Firstly, we classified the participants into three groups

according to their answers to the post-experiment

questionnaire. Nine participants whose answers were

7, 6, 5 (in the 7-point Likert scale) to the questions of

"I regarded the avatar as my alter-ego." are classified

as "high group," 5 participants as "middle group"

(score=4), and 8 as "low group" (scores 3, 2, 1). The

following analyses used scores of the high group and

the low group only in order to compare the answers of

the opposite groups. We conducted a 2-way ANOVA

(repeated measures) to the answers of BIGFIVE

extroversion questions. The factors are "avatar

(combination of the attractiveness of the self-avatar

and the accompanying avatar (AT/AT, AT/NAT,

NAT/AT, NAT/NAT) " and "participant group

(high/low)." In addition, a one-factor analysis of

variance was conducted on the extroversion answers

of the pre-experiment questionnaire and the main

questionnaire.

There are significant main effects and interactions

on the evaluation of extroversion in terms of sociable,

talkative, cheerful, amiable, outgoing, assertive, and

energetic. We show preliminary results of sociable,

cheerful and amiable below.

The evaluation result of “sociable” is shown in

Figure 6. The main effect was observed in the

participant group conditions. The high group’s

sociable evaluation was significantly higher than that

of the low group (Score=4.8-3.1, F=11.13, p≤0.01).

The result of multiple comparisons showed that the

high group’s sociable evaluation scores were

significantly higher than those of the low group under

4 avatar conditions made in the virtual world

(AT/AT: F=16.76, p≤0.01, AT/NAT: F=13.77,

p≤0.01, NAT/AT: F=8.5, p≤0.05, NAT/NAT: F=4.56,

p≤0.05). However, there was no significant difference

in the sociable evaluation between the real and virtual

conditions. High group’s sociable evaluation of the

NAT/AT condition was significantly lower (F=4.66,

p≤0.05) than that of the AT/AT condition, and similar

tendency was shown between the AT/AT and the

NAT/NAT condition (F= 4.29, p≤0.1). On the other

hand, there was no significant difference in the

evaluation among the avatar conditions in the low

group.

The evaluation result of “cheerful” is shown in

Figure 7. The main effect was observed in the

participant group conditions. The high group’s

cheerful evaluation was significantly higher than that

of the low group (F=7.74, p≤0.05). The result of

multiple comparisons showed that the cheerful

evaluations of the high group were higher in the

AT/AT (F=7.09, p≤0.05), the AT/NAT (F=9.71,

p≤0.01), and the NAT/AT condition (F=4.76, p≤0.05)

than those made by the low group. The high group’s

cheerful evaluation of the NAT/NAT condition was

significantly lower (F=6.28, p≤0.05) than that of the

AT/AT condition, while there was no significant

difference in the evaluation among the avatar

conditions in the low group.

The evaluation result of “amiable” is shown in

Figure 8. The main effect was observed in the

participant group conditions. The high group’s

amiable evaluation was significantly higher than that

of the low group (Score=5.3-3.9, F=6.12, p≤0.05).

The result of multiple comparisons showed that the

high group’s amiable evaluation scores were

significantly higher in the AT/AT (F=10.24, p≤0.05),

and the NAT/NAT condition (F= 10.5, p≤0.05) than

those made by the low group. In addition, the high

group’s evaluation of their amiableness in the AT/AT

(Score=4.6-5.6, p=0.07) and the NAT/NAT

(Score=4.6-5.6, p≤0.05) conditions were significantly

(in NAT/NAT) higher than their amiable evaluations

in the real world. On the other hand, there was no

significant difference in the evaluation among the

avatar conditions in the low group.

Figure 6: Comparison of sociable evaluation.

Figure 7: Comparison of cheerful evaluation.

Analysis of the Effects of Appearances of Avatars on User’s Self-evaluation of Extroversion

235

Figure 8: Comparison of amiable evaluation.

4.2 Discussion

In term of the sociable evaluation, comparison

between the participant groups and the avatar

conditions showed that the sociable evaluation

scores of the high group were significantly higher

than those made by the low group in all four avatar

conditions. In addition, the sociable evaluation score

of the NAT/AT and the NAT/NAT condition were

lower than that of the AT/AT condition in the high

group. This suggests the possibility that the high

group’s evaluation of their sociable-ness in the

virtual world was affected by the attractiveness of

their self-avatar, regardless of the attractiveness of

their accompanying avatar. On the other hand, the

appearance of the avatars did not influence their

sociable evaluation scores in the low group, where

their scores were always lower than the high group,

and were always lower than the median value.

In terms of the cheerful evaluation, the high

group’s evaluation scores were significantly higher

in the three avatar conditions than those of the low

group. The high group’s cheerful evaluation of the

AT/AT condition was significantly lower (F=6.28,

p≤0.05) than that of the NAT/NAT condition, while

there was no significant difference in the evaluations

among the avatar conditions in the low group. This

suggests that the high group evaluated their

cheerfulness higher when their self-avatar's

appearance was attractive than they did when their

self-avatar was not attractive. However, in the low

group, no significant difference was observed in all

avatar conditions and the evaluation scores were

lower than the median value. Therefore, similar to

the sociable evaluation, the appearance of avatars

did not affect the low group’s low self-evaluation of

cheerfulness, where their scores were always lower

than the high group.

In terms of the amiable evaluation, the

evaluation of the AT/AT and the NAT/NAT

conditions were significantly higher in the high

group than those in the low group. In addition, the

high group’s evaluation of their amiableness in the

AT/AT and the NAT/NAT conditions were

significantly higher than their amiable evaluation in

Yee and Bailenson the real world. This suggests that

when the attractiveness of the self-avatar and the

accompanying avatar is balanced, the high group

evaluated themselves as more amiable than they did

in other conditions where the attractiveness of the

avatars are not balanced. On the other hand, there

was no significant difference in the evaluation

among the avatar conditions in the low group. This

result suggests that the low group has low appraisal

of self's amiableness regardless of the avatar's

appearances.

Those results suggest there is a dichotomy

between the participant groups in terms of self-

assessment of extroversion. The tendency is that the

extroversion evaluation of self is influenced by the

appearance of self-avatars in the high group,

although not all evaluations of BIGFIVE

extroversion scores rose in the conditions when

attractive self-avatars were used. Even when the

appearance of the self-avatar is not attractive, if the

attractiveness of the self-avatar and the

accompanying avatar is balanced, the extroversion

evaluation of self is influenced in the high group.

Thus, part of the hypothesis 1 "Those who

regard their self-avatar as their own alter-ego, if their

self-avatar and accompanying avatar are attractive,

their self-evaluation of extroversion becomes higher

compared to when the avatars are not attractive."

was partially supported. In other words, the high

group has a tendency of evaluating their self-

extroversion higher when their self-avatar is

attractive. Their extroversion evaluation is also

affected by the combination of self and

accompanying avatar’s attractiveness, regardless of

the attractiveness of the accompanying avatar alone.

On the other hand, no significant difference was

observed in the evaluation scores of the BIGFIVE

extroversion in the low group. This suggests that

extroversion evaluation of the low group is not

affected by the appearance of the avatars. Thus,

Hypothesis 2: "Those who do not regard their self-

avatar as their alter-ego, their self-evaluation of

extraversion does not change regardless of the self-

avatar's or accompanying avatar's appearances." was

supported. Compared with the high group, the low

group has a low BIGFIVE extroversion evaluation in

real world, and their evaluation is constantly low

regardless of the appearance of the avatars in the

virtual space. This result can be explained that the

ICAART 2019 - 11th International Conference on Agents and Artiﬁcial Intelligence

236

low group did not regard their self-avatar as their

alter ego, so the avatars’ appearances does not affect

their low extroversion evaluation in the virtual

world, which is also low in the real world.

Future directions of this research are as follows;

it is necessary to classify the reason as to whether

the avatars are regarded as their alter-ego by user

attributes, i.e., by personality test result other than

extroversion, frequency of playing online games,

years of online game experience, etc. Moreover, we

should consider gender effect in more details,

namely, we should examine the Proteus effect by

considering the gender of the participants more

precisely. We had only Japanese participants in the

current study. However, we should consider other

cultures’ preferences on avatar and agents

appearances.

5 CONCLUSION

In this research, we examined the influence of the

appearance of the self and accompanying avatars on

the participants’ extroversion evaluation in the

virtual world. The participants were divided into two

groups depending on whether they regarded the self-

avatar as their alter ego. The result indicated that

there is a dichotomy between the participant groups

in terms of self-assessment of extroversion. The high

group (who regarded the avatar as their alter-ego)’s

extroversion evaluation was affected by the

appearance of the self-avatar, thus Proteus effect

was applicable. On the contrary, the group that did

not regard the self-avatar as their alter ego did not

change their low extroversion evaluation according

to the appearance of their self-avatar nor their

accompanying avatar. The results suggest the

importance of considering avatars’ and agents’

appearances and user’s personality in serious games,

online virtual worlds, and avatar-mediated online

communications.

ACKNOWLEDGEMENTS

This work was partially supported by JSPS

KAKENHI, JP26330236, and JP17K00287.

REFERENCES

Gough, H.G, and Heilbrum, A.B. 1983. The Adjective

Check List Manual. 1983ed. Palp Alto, CA.

Consulting Psychologist Press.

Pan Y, and Steed A. 2017. The impact of self-avatars on

trust and collaboration in shared virtual environments.

PLoS ONE 12(12): e0189078.

https://doi.org/10.1371/journal.pone.0189078.

Peña, J., Hancock, J.T., and Merola, N.A. 2009. The

Priming Effects of Avatars in Virtual Settings.

Communication Research, 36(6), 838-856.

Takeuchi, K., Hasegawa, D., and Sakuta, H. 2016.

The Effects of Space Representation on Dialog

Content in a 3D Avatar Chat System. IPSJ

Interaction2016, 746-750.

Thaler, A. et al. 2018. Visual Perception and Evaluation of

Photo-Realistic Self-Avatars From 3D Body Scans in

Males and Females. Frontiers in ICT, 5,

https://doi.org/10.3389/fict.2018.00018.

Tokioka, R. 2014. A Clinical Psychological Study of the

“Other Self” in a Virtual Space, Kyoto University

research studies in education, 60, 235-247.

Yee, N. and Bailenson, J. N. 2007. The Proteus Effect:

The Effect of Transformed Self-Representation on

Behavior. Human Communication Research, 33, 271-

290.

Yee, N., Bailenson, J. N., and Ducheneaut, N. 2009. The

Proteus Effect: Implications of Transformed Digital

Self-Representation on Online and Offline Behavior.

Communication Research, 36(2), 285-312.

Analysis of the Effects of Appearances of Avatars on User’s Self-evaluation of Extroversion

237