مال من، خودم و من: نقش زمینه تعاملی در خودارائه گری از طریق آواتار

Abstract This paper investigates whether the nature of an online environment can prime users to create avatars that emphasize particular characteristics. Participants created an avatar for one of three contrasting settings: blogging, dating or gaming. For the most part, avatars in blogging were created to accurately reflect their owners’ physical appearance, lifestyle and preferences. By contrast, participants in the dating and gaming treatments accentuated certain aspects of their avatar to reflect the tone and perceived expectations of the context. For instance, avatars in dating were made to look more attractive while avatars in gaming were made to look more intellectual. Yet, predominantly, these emphasized avatar attributes drew on participants’ self-image, and thus avatars were perceived by their owners as highly similar to themselves. The implications of these results are discussed against current frameworks of online identity and behavior. Most importantly, we use our results to extract design recommendations for improving avatar-driven applications.

1. Introduction This paper presents an experiment that investigates how users customize avatars for self-presentation purposes in the following three online settings: blogging, dating and gaming. We go on to consider whether self-presentation as chosen for the three settings impacts on users’ perceptions of similarity with their avatar and in turn, on self-awareness. Avatars present a creative platform for identity construction in computer-mediated communication (CMC). While avatars maintain users’ privacy when needed, they give expressive freedom over an otherwise anonymous and static online presence. When constructing avatars, users take care to create unique representations despite available ready-made options (Cheng et al., 2002 and Taylor, 2002). This has been found to be the case even when the customization process imposed by the system was exceptionally time demanding (Cheng et al., 2002). Designers of social systems have recognized the significance of avatars for identity expression by integrating them across a variety of online applications including virtual environments, gaming and blogging. As avatars begin to replace other conventional forms of visual identity, e.g., photographs, it becomes important to understand whether an avatar can alter users’ perception and subsequent behavior during online communication. To that end, the majority of research on avatars has focused on how an avatar’s visual realism and behavioral realism can impact on the viewer’s perception (e.g., Garau et al., 2003, Nowak and Rauh, 2005 and Nowak and Rauh, 2008). Likewise, an avatar’s appearance can influence its owner’s perception and behavior. For instance, users, who perceive their avatar as more similar to their own appearance, become more self-aware (Vasalou, Joinson, & Pitt, 2007) a state that may impact on the quality of online communication by encouraging higher rates of self-disclosure (Joinson, 2001). Yee and Bailenson (2007) found that qualities displayed in an avatar’s appearance can steer its owner’s behavior. Participants who were assigned to more attractive avatars demonstrated increased self-disclosure and willingness to approach members of the opposite gender than participants represented by less attractive avatars. In a negotiation task, participants with taller avatars were more confident than participants represented by shorter avatars. In regards to self-presentation, it is not known what type of environment might lead online users to create avatars that highlight particular characteristics. Are users of online dating more likely to choose attractive avatars than users of online gaming? As concerns perception, are attractive avatars perceived by users to be equally self-representative as compared to avatars that highlight more strategic characteristics? These open questions make it difficult to relate the results of the above-mentioned perception studies to particular online situations. Moreover, understanding how users choose to present themselves for certain settings can provide realistic scenarios for designing future perception studies. This information can also offer design recommendations for avatar customization tools. If designers are mindful of the customization options preferred by users of a specific domain, these options can be highlighted and made easily accessible in the avatar customization interface. In the present research, we experimentally manipulated the context of avatar use. Users created an avatar with the intention of participating in a blogging, dating or gaming environment. A first aim of the research was to reveal how users choose to present themselves through an avatar for each of these three contrasting settings. A second objective was to discover in which setting avatars are perceived as self-representative, and in consequence which of the three environments elicits higher self-awareness. We first position these two objectives against existing views on self-presentation and self-awareness; this background review is used to formulate several hypotheses and research questions. We then detail the experimental study we carried out, which as described above, comprised three conditions: blogging, dating and gaming. Our two-tier quantitative and qualitative methodological approach is also presented. An analysis of the results follows which reveals that participants by and large created self-representative avatars irrespective of their context of use. Nonetheless, avatars were slightly morphed to align to the expectations of blogging, dating and gaming. As participants predominantly reported perceiving their avatars as similar to themselves, reports of self-awareness did not differ across the three experimental conditions. A discussion follows that elaborates on these findings and their implications for the design of avatar customization tools.

. Results Section 5.1 examines how participants presented themselves via their avatar by testing hypothesis 1 and 2 as formulated in Section 2. Section 5.2 explores participants’ perceived similarity with their avatar in the three conditions and whether perceptions of similarity influenced state self-awareness. Section 5.3 presents the qualitative results of the post-interviews. 5.1. Self-presentation The first analysis addressed hypotheses 1 and 2. Earlier, we postulated that participants of the dating condition will slightly exaggerate physical and personality attributes as compared to blogging participants (H1). The physical attributes captured by the AVATAR ATTRIBUTES measure is physical attractiveness. The personality attributes captured by the AVATAR ATTRIBUTES measure are intellectual ability, social skills/social competence, artistic and/or musical ability, athletic ability, leadership ability, common sense, emotional stability, sense of humor and discipline. Participants of the gaming condition were expected to exaggerate strategic attributes as compared to blogging participants (H2). The strategic attributes captured by the AVATAR ATTRIBUTES measure are intellectual ability, athletic ability, leadership, common sense and discipline. A multivariate ANOVA was conducted with condition (blog, date, game) as a factor and the 10 AVATAR ATTRIBUTES as the dependent measures. Two planned contrasts were performed by using the blog condition as the reference category (−1 +1). Overall main effects for condition and main effects for the 10 dependent variables were non-significant. We report the significant results of the planned contrast at the .05 alpha cut off. Self-presentation attributes as reported for the avatar are shown in Table 1. Table 1. Descriptive statistics. Avatar attributes Self-attributes Computed z-scores Mean SD Mean SD Mean SD Blogging Intellectual ability 7.32 2.079 8.45 .912 −.802 1.698 Social skills/social competence 7.18 1.868 6.95 2.380 .082 1.017 Artistic and/or musical ability 6.45 1.765 6.00 1.927 .348 .884 Athletic ability 6.27 2.229 5.73 1.980 .290 1.194 Physical attractiveness 6.41 2.153 6.45 1.792 −.158 1.612 Leadership ability 6.59 1.817 7.27 1.352 −.243 1.166 Common sense 7.09 2.180 7.64 1.891 −.306 1.340 Emotional stability 6.82 2.062 7.23 1.950 −.130 1.064 Sense of humor 7.45 1.371 7.86 1.390 −.108 .952 Discipline 6.45 1.969 6.66 2.415 −.230 .898 Dating Intellectual ability 7.68 1.376 8.40 1.354 −.506 1.124 Social skills/social competence 7.40 1.472 7.12 1.616 .201 .801 Artistic and/or musical ability 6.76 1.899 5.76 2.146 .501 .951 Athletic ability 6.20 1.756 6.04 1.695 .251 .940 Physical attractiveness 7.28 1.208 6.84 1.179 .494 .905 Leadership ability 6.60 1.732 6.76 1.562 −.237 1.111 Common sense 7.40 1.258 7.88 1.394 −.116 .773 Emotional stability 6.80 2.236 6.88 2.027 −.139 1.154 Sense of humor 7.20 1.443 7.60 1.291 −.284 1.003 Discipline 6.72 1.696 7.40 2.062 −.109 .774 Gaming Intellectual ability 7.75 1.751 8.04 1.334 −.449 1.430 Social skills/social competence 7.38 2.081 7.00 1.532 .187 1.133 Artistic and/or musical ability 5.96 2.010 5.54 1.956 .099 1.007 Athletic ability 5.96 2.216 5.42 1.954 .122 1.187 Physical attractiveness 6.92 1.840 6.54 .977 .222 1.377 Leadership ability 7.08 2.104 6.92 1.742 .072 1.350 Common sense 7.17 1.993 7.25 1.595 −.260 1.225 Emotional stability 6.58 2.466 7.13 1.895 −.251 1.272 Sense of humor 7.33 1.810 7.38 1.637 −.192 1.257 Discipline 6.88 2.473 6.79 2.126 −.038 1.128 Table options Participants in the dating condition as opposed to those in the blogging condition, reported more “physical attractiveness” (F(2, 68) = 1.44; p < .05). Thus, contrary to the first hypothesis, participants of dating did not inflate the entire spectrum of avatar attributes but only accentuated the avatar’s physical attributes. In the gaming condition, none of the measures reached significance when contrasted to the blogging condition. The second hypothesis was not confirmed. Therefore, the results of the planned contrasts showed hardly any differences between the three conditions, with the exception of “physical attractiveness” in the dating condition. Hypotheses 1 and 2 were re-examined in a second analysis; the previous analysis was extended by determining whether the avatars created in the three conditions differed when accounting for the population’s self-attributes.1 A series of z-scores computed the difference between a given avatar attribute and the population’s report on the same attribute, but as given for self. For example, when calculating the score for intelligence, in z = (X − μ)/σ the mean and standard deviation for the self-attribute was taken from the entire sample of 71 participants. X was the value reported by each participant for avatar intelligence. Self-presentation attributes as reported for self as well as the computed z-scores are reported in Table 1. A negative value means that the avatar was, on average, rated lower than the self. A positive score indicates that the avatar was rated higher on the attribute than the self. The multivariate analysis was performed again but with the 10 z-scores as dependent variables. As in the first analysis, planned contrasts compared blogging to dating and gaming. The results from the first analysis were replicated. Physical attractiveness in dating was significant (F(2, 68) = 1.44; p < .05) as compared to blogging. 5.2. Perceptions of similarity and self-awareness The second analysis addressed whether participants’ perception of similarity with their avatar changes as a result of the different self-presentation strategies used in the three online environments, and in consequence whether perceived similarity impacts on state self-awareness. At a first glance, the results from the previous Section seem to suggest that participants will perceive their avatar similar to themselves irrespective of the context. However, there may be other factors that alter similarity perceptions which were not captured by the AVATAR ATTRIBUTES and SELF-ATTRIBUTES measures. Therefore, this analysis is still important to pursue. A one-way way ANOVA was first performed with the dependent variable SIMILARITY and condition as a factor. No prior hypotheses were made for similarity reports; therefore post hoc tests were chosen using the Tukey criterion. The main effect was non-significant (F(2, 68) = 1.35; p > .05) while none of the comparisons between conditions yielded a significant result. There was some non-normality in the data. To address this, a Kruskal–Wallis non-parametric test was also performed and showed a non-significant main effect x2 = 3.84, df = 2, p > .05. Fig. 1 displays the distribution of similarity ratings in the three conditions and in the entire sample. It is evident that most participants reported high similarity with their avatar. Out of 71 participants, 82% were above the mean of the scale, thus the sample is skewed towards the highest scores of similarity. Frequency distribution of reported similarity. Fig. 1. Frequency distribution of reported similarity. Figure options The similarity reports found in (Vasalou et al., 2007) followed a normal distribution. This enabled an associative analysis that correlated participants’ reported SIMILARITY with STATE SELF-AWARENESS. Here, given the low variance in the similarity scores, it is unlikely we will find a positive linear relationship between SIMILARITY and STATE SELF-AWARENESS. A scatterplot showed there was no relationship between SIMILARITY and STATE SELF-AWARENESS; the slope was close to zero and the intercept was 2.10. Therefore, a correlation analysis was not pursued.2 5.3. Post-interviews 5.3.1. Blogging In the blogging condition, many participants reported mapping their personality and preferences onto their avatar. For example, one participant placed her avatar in a cabin in the woods where she wishes to live someday. A golden retriever was selected which is the type of dog she hopes to own in the future. The avatar’s style of clothes corresponded to her taste in fashion (Fig. 2b). A second participant told us that playing “Nintendo Wii” is his favorite activity. He thus selected a robot and a gaming background suggestive of his hobby. His avatar’s physical appearance was chosen to reflect his own appearance: a shirt with rolled-up sleeves followed his style of dressing and the avatar’s happy expression denoted his positive attitude (Fig. 2c). A third participant placed his avatar against a London background, his city of residence, with his favorite pet, a small dog (Fig. 2d). (Blog) (a) I watched a Sci-Fi movie last night, (b) My desired lifestyle, (c) I ... Fig. 2. (Blog) (a) I watched a Sci-Fi movie last night, (b) My desired lifestyle, (c) I love to play games, and (d) This where I live. Figure options When analyzing the interviews, it became apparent that the notion of “accurate” self-representation that participants repeatedly evoked included three key features. Firstly, stable aspects of oneself were mirrored onto the avatar. Participants opted for choices they liked, such as fashion items they often wear offline. Hobbies, such as playing games or an interest in history, were displayed. Secondly, participants chose options that mirrored upcoming, past or current life happenings, thus expressing anticipated events or precious memories. Thirdly, desired life changes and self-improvements were also reflected. A few participants chose pets they would like to own someday. One participant expressed the carefree lifestyle she is striving for. Another participant reflected his lifelong dream to visit New York with the corresponding background option. Although accurate self-representations were predominant, four out of 22 participants constructed avatars that combined few real life aspects with many fantasy options. One participant began by choosing an Egyptian top for his avatar because of his interest in ancient Egypt. However, the previous night he had watched a Sci-Fi movie whose movie plot he recreated by constructing a fantasy avatar (Fig. 2a). In this example, a real life event was referenced to create a fictional character. In the remaining three instances, participants used avatars as a communication tool. Options were taken because they were amusing and entertaining to them while participants imagined their friends’ reactions in response to the blog entry. 5.3.2. Dating In the dating condition, participants repeatedly verbalized that avatars were created to impress by displaying their romantic, attractive, or intellectual qualities. Yet, the choices participants took, reflected their taste, preferences and lifestyle. For instance, three participants chose an autumn background, because of its “soft and romantic” colors (Fig. 3a). After selecting this background, one participant noted that this was a place he could imagine going to. Another participant selected a backdrop of her favorite type of restaurant for a first date (Fig. 3c). At the same time, she wore clothes similar to those she was wearing the previous day. The interplay between one’s desire to express his/her stable qualities while also wanting to impress becomes clearer in the following example. A male participant first selected a summer background and summer clothes to reflect the season. Midway through the avatar creation, he came across a ballroom dancing background. At that time, he remembered the dating context; he chose this new background because it was more appropriate and because he practices ballroom dancing (Fig. 3d). Hence, in our view, although the dating context led many participants to inflate qualities required to gain attention from the opposite gender, the choices they used to impress were congruent with their personality. (Date) (a) I am romantic, (b) My avatar is attractive, (c) My favorite ... Fig. 3. (Date) (a) I am romantic, (b) My avatar is attractive, (c) My favorite restaurant for a date, and (d) I practice ballroom dancing. Figure options Despite this predominant trend, five out of 25 participants applied two additional (albeit less salient) self-presentation strategies. Firstly, two female participants created deceptive avatars aiming to manipulate the viewer’s perception. For example, one participant composed a “hot beach girl”. Even though she was a brunette, the avatar’s hair was changed to red as it rendered a more beautiful image for the viewer. She deliberately chose a sexy short skirt and a sleeveless shirt despite her own conservative fashion style (Fig. 3b). Secondly, three participants maintained that their avatar reflected themselves and was not modified any differently for this setting. These three participants drew on “accurate” self-presentation strategies as those found in the blogging condition. 5.3.3. Gaming In the gaming condition, participants developed four very diverse strategies to depict themselves. Firstly, similarly to the dating condition, here also, nine participants maintained that the avatar was not altered in any particular way for the game. These participants used “accurate” self-presentation strategies as was observed in the blogging condition. Secondly, six participants created avatars that were self-representative but yet slightly “smarter”, “more creative”, “sporty”, “cooler” or “capable” because of the game. Fig. 4c displays an avatar that was constructed by its owner to look wittier. While applying this second strategy, a few participants experienced a feeling of drifting from their real self towards designing an avatar specifically for the game. In the course of this negotiation process, a female participant chose a piano depicting her musical hobby, her favorite hairstyle and her college books, but with a slight fantasy alteration, butterfly wings, due to the game (Fig. 4d). (Game) (a) I am a suspect, (b) I am a detective, (c) I look witty, and (d) This ... Fig. 4. (Game) (a) I am a suspect, (b) I am a detective, (c) I look witty, and (d) This is who I am. Figure options In a further, third approach eight participants morphed their choices in accordance to the visual style and type of the game. Aligned to the visual style of the game, one participant chose a dark palace background as he expected the game to look three-dimensional. Another participant chose a street background so that “it blends in the game”. Aligned to the type of the game, six participants engaged in role playing. The application of this strategy was variable, ranging from a subtle orientation towards a particular role, to role playing with one’s entire embodiment. For example, while two participants constructed self-representative avatars that emitted subtle visual qualities of a detective, e.g., thick black eyewear, four others created avatars whose entire appearance, e.g., clothes, accessories and background, rendered the detective’s image (Fig. 4b). In a fourth approach, two participants selected customization options to distract other players as a game strategy. A male participant created a “sinister” female avatar. This was done so that other players in the game would focus their attention on the character, thus allowing him to win the game (Fig. 4a). Likewise, a second male participant chose a swirling background that he believed was hypnotic and distractive to other players.