- Email: [email protected]
The effects of static avatars on impression formation across different contexts on social networking sites
Computers in Human Behavior 53 (2015) 111–117
Contents lists available at ScienceDirect
Computers in Human Behavior journal homepage: www.elsevier.com/locate/comphumbeh
Full Length Article
The effects of static avatars on impression formation across different contexts on social networking sites David Westerman a,⇑, Ron Tamborini b, Nicholas David Bowman c a
North Dakota State University, Department of Communication, Department #2310, P.O. Box 6050, Fargo, ND 58108, United States Michigan State University, Department of Communication, 404 Wilson Road, Room 570, Communication Arts and Sciences Building, East Lansing, MI 48824, United States c West Virginia University, Department of Communication Studies, 108 Armstrong Hall, P.O. Box 6293, Morgantown, WV 26506, United States b
a r t i c l e
i n f o
Article history: Received 2 December 2014 Revised 10 June 2015 Accepted 12 June 2015 Available online 7 July 2015 Keywords: Avatars Uncertainty Attractiveness Impression formation
a b s t r a c t When making judgments about others, people use whatever social information is available in online environments. Such is the case for forming impressions of others. One type of such social information is a user’s avatar. This study examines different types of avatars (photographs, cartoon humans, and nonhumans) created for task, social or dating/romantic situations to study the effect of avatar type on judgments of uncertainty and task-specific attractiveness. Data suggest various patterns of uncertainty and attractiveness in these situations. Both the graphic from of an avatar and the context of impression formation have effects on subsequent impression formation. Judgments of uncertainty and attraction were affected by both the graphic from of avatar and by the consistency between the context of impression formation and the attractiveness cues of the avatar. These findings are discussed as are implications for future research. Ó 2015 Elsevier Ltd. All rights reserved.
1. Introduction
1.1. The use of visual information (avatars) in SNS
Social information processing theory (SIPT; Walther, 1992) assumes that people want to (and can) enact interpersonal processes online in order to attain interpersonal goals. One goal that is central to communicators – online or offline – is reducing uncertainty about others (Berger & Calabrese, 1975). In the text-based systems that typified CMC at the time of SIPT’s inception, people had few, if any, visual cues available to use for uncertainty reduction purposes. However, SIPT also suggests that people use the information that is available in an environment to accomplish their goals. Thus, even as more modern CMC systems provide increased visual cues – such as Facebook’s recent redesigns that emphasize visual information (Rodriguez, 2013), SIPT provides a useful framework to explain how users accomplish interpersonal goals online. The current study examined how increased visual information (of even a relatively low level) in a social networking site (SNS) can impact the impression formation process online.
Although some CMC is done through text-only channels (Walther & Parks, 2002), there are growing areas of CMC that incorporate visual information about communicators. One prominent online space utilizing such information is the social networking site (SNS). These sites provide cue-rich arenas for users to communicate using a mix of textual and visual cues. In fact, one of the hallmarks of such sites is that they allow people to ‘‘construct a public or semi-public profile’’ (boyd & Ellison, 2007, p. 211) and part of that profile construction is the inclusion of pictures. For example, Facebook, the largest SNS with over one billion monthly active users as of March 2014 (Facebook, 2014), recently reported that over 350 million unique photos are uploaded to its servers every day (Kotenko, 2013). One important piece of visual information that SNS users will provide is how they choose to represent themselves in their profile pictures. Visual appearance plays a big role in the impression formation process, both online and offline. This is especially true during first impressions of strangers, when nonverbal information can lead to spontaneous impressions of another person within a few seconds (Schneider, Hastorf, & Ellsworth, 1979) that can be very resistant to change (Kelley, 1950). Content analytic work by Hum et al. (2011) found that the majority of a sample of 150 college student Facebook profile pictures tended to be posed, inactive (not in motion), and containing only the subject. However, a profile
⇑ Corresponding author. E-mail addresses: [email protected] (D. Westerman), tamborin@ msu.edu (R. Tamborini), [email protected] (N.D. Bowman). http://dx.doi.org/10.1016/j.chb.2015.06.026 0747-5632/Ó 2015 Elsevier Ltd. All rights reserved.
112
D. Westerman et al. / Computers in Human Behavior 53 (2015) 111–117
picture does not have to be a photograph of the person. Instead, it can be what might be more generally referred to as an avatar. An avatar can be defined as a users’ graphical representations in a given virtual environment (Nowak, 2000), which can also include simple icons chosen as a form of self-representation (Suler, 1997). Overall, avatars can range from simple, static images to more animated and dynamic characters that are chosen to represent a person online (Bailenson & Blascovich, 2004). Avatars are different from agents in that avatars represent an actual person, whereas agents are computer-controlled entities. These various types of representations have a long history in online spaces, dating back at least to the use of icons, photographs, and other symbols in personal home pages (Chandler, 1998). Online spaces, such as SNS, provide users an increased opportunity to control their self-presentation (Walther, 1996). For example, people tend to exaggerate their attractiveness in their dating profiles, reporting the most deception in their profile pictures (Toma, Hancock, & Ellison, 2008). Donath (2007) also points out that people infer various character traits about other people based upon the visual appearance of an avatar. Van Der Heide, D’Angelo and Schumaker (2012) found that users privilege the photographic over textual information on Facebook profiles. Utz (2010) found that people were viewed as more socially attractive when their profile was more ‘‘extraverted’’, including a more extraverted picture. Friends’ profile pictures also had an impact on judgments of the profile owner (Utz, 2010; Walther, Van Der Heide, Kim, Westerman, & Tong, 2008). Notably, these are photographs used to represent people in various social profiles. What is unknown is how people will respond to various types of avatars that one might use to represent oneself in SNS. Thus, the following two research questions are offered: RQ1: How does the graphic form of an avatar (photograph, cartoon human, and nonhuman) influence a receiver’s uncertainty regarding the source? RQ2: How does the graphic form of an avatar (photograph, cartoon-human, and nonhuman) influence receiver perceptions of source attractiveness (social, physical, and task attractiveness)? 1.2. The role of context in judgments of people using avatars Past research shows that people make judgments based upon the avatars used in various platforms. These perceptions include intelligence (Koda, 2004), sociability and attractiveness (Weibel, Stricker, Wissmath, & Mast, 2010), personality traits (Marcus, Machilek, & Schütz, 2006), uncertainty (Nowak, 2004), credibility (Nowak & Rauh, 2005, 2008), group identity (Kim, 2009; Kim & Park, 2011; Lee; 2004; Lee & Nass, 2002) and affiliation (Lortie & Guitton, 2011; 2012). In general, the research suggests that people make judgments of avatars and the people that use them. Although people may make judgments of others based on visual cues, this is not the only source of information that may be used. Tagiuri (1969) suggests that people form impressions of others by a combination of object cues and situational context. In fact, past studies involving avatars in CMC (Nowak & Rauh, 2005) have called for more research examining the effects of different contexts on impression formation. We might expect that avatars created to provide information that was consistent with goals of the interaction context would do a better job of reducing uncertainty and increasing liking. Why might this be the case? Generally stated, information that conforms to expectations should reduce uncertainty, whereas an expectancy violation may increase it. Expectancy violations tend to increase uncertainty (Berger, 1993; Planalp & Honeycutt, 1985). Applied to CMC, an avatar that fits preconceived notions
of an interaction context should reduce uncertainty, whereas one that violates preconceived notions of what to expect in a given context may raise questions and increase uncertainty. In FtF settings, an ambiguous interaction context has been found to create more uncertainty than a more specific, task-based context (Rubin, 1977). This suggests that context of an interaction can reduce uncertainty by providing a focal point for initial interaction. In other terms, context helps specify which information in an interaction is important to a task (Kelly, 1955). Without an understood context for interaction, it is difficult to imagine what information participants in previous research might have gained from avatars to help reduce uncertainty or form initial impressions about their interactional partner. At a broader level, the basic information one usually provides on a social network – starting with the earliest MySpace pages to more popular Facebook profiles – may be thought of in terms of McCroskey and McCain’s (1974) classic measure of interpersonal attraction: namely, social attraction, physical attraction, and task attraction. Most SNSs, even sites more aligned with professional rather than personal networking such as LinkedIN, request one to provide their physical picture (physical attractiveness) as well as their hobbies and interests (social attractiveness) as well as their education and work experience (task attractiveness). The focus on these features of social interaction points to their importance in interpersonal communication and calls our attention to them in efforts to understand online communication. We presume that different expectations for an interaction are established by the belief that an online exchange is motivated by one purpose versus another. Moreover, we should anticipate that uncertainty reduction is governed by the extent to which information received during an online exchange conforms to expectations created by these different contexts. Information that conforms to expectations should reduce uncertainty, whereas an expectancy violation may increase it. In this manner, information interacts with context to influence uncertainty reduction. For example, cues that conform to expectations for a social interaction task might conflict with expectations for a business related task. This leads to the following hypotheses: H1. The context of an online interaction (potential dating, hanging out, or task achievement) interacts with cues present in an online avatar (physical, social, and task attractiveness) to influence perceptions of uncertainty regarding the source such that cues that conflict with expectations for the interaction context lead to greater uncertainty regarding the source than cues that conform to those expectations.
H2. The context of an online interaction (potential dating, hanging out, or task achievement) interacts with cues present in an online avatar (physical, social, and task attractiveness) to influence perceptions of source attractiveness such that cues that conform to expectations for the interaction context lead to greater source attractiveness than cues that conflict with those expectations.
2. Method 2.1. Overview A 3 3 mixed design varied the context given to respondents for online interaction and the attractiveness cues of avatars attributed to 15 apparent online partners. The between subjects factor was context. Respondents were randomly assigned to one of three conditions varying the reason given for their online interaction (for potential dating, hanging out, or task achievement). The within
D. Westerman et al. / Computers in Human Behavior 53 (2015) 111–117
subject factor varied three dimensions of attractiveness (physical, social, and task attractiveness) in the avatars by manipulating each avatar related to physical good looks, social gregariousness, and task achievement. Respondents were asked to evaluate those individuals online who were purported to have chosen each of the 15 different avatars using scales designed to measure three dimensions of uncertainty and attractiveness. 2.2. Participants 206 undergraduates enrolled in introductory communication courses at a large Midwestern university participated in this study in exchange for course credit. The sample ranged in age from 17 to 24 (M = 19.95, SD = 1.30). Most of the sample (80.1%) was female with three people (1.5%) not responding. Racially, the sample was comprised of 22 (10.7%) African-Americans, five (2.4%) Asian-Americans, 164 (79.6%) Caucasians, four (1.9%) Hispanics, three (1.5%) mixed, and seven (3.4%) others; one person (0.5%) did not respond. 2.3. Procedure Participants were told that they would be offering comments on profiles created by students ‘‘in another class.’’ They were seated together in groups ranging in size from seven to 13, but did not interact with each other. After assignment to one of the three interaction settings, respondents were given instructions that primed attention to the profile creators’ physical, social, or task attractiveness. The instructions for the dating condition (physical attractiveness) were as follows: ‘‘Thank you for your willingness to participate in this study. Today, you are going to be asked to participate in a study similar to speed dating, only online. Students in another class were asked to create profiles for this speed dating. You are about to see ten of these profiles. After you have viewed all 10 profiles, you will be asked to rate each person for how much you would potentially like to go on a date with them. Please note that you are being randomly assigned profiles from this class, so you may see profiles of both males and females. After you rate each person, you will be asked to answer some questions about each one, and then some questions about yourself. At the end of the entire study, four couples will be randomly chosen and will be offered the chance to go on a movie date. Those couples will be chosen by taking the people with the four highest rated profiles, and randomly selecting one person that had them ranked as their top choice.’’ The instructions for the task achievement condition (task attractiveness) and the hanging-out condition (social attractiveness) were similar to those for the dating condition with the following exceptions: First, instead of respondents being asked to ‘‘participate in a study similar to speed dating,’’ in the task condition they were asked to ‘‘participate in a study that may give you a chance to work with a local movie theater in promoting a local film that will be playing in January,’’ and in the hanging-out condition they were asked to ‘‘choose a group of people you think would be fun to hang out with.’’ Also, instead of being asked to rate how much they would like ‘‘to go on a date with’’ the other person they were asked to rate how much they would like ‘‘to work with them’’ or ‘‘to hang out with them.’’ Finally, instead of being offered the chance ‘‘to go on a movie date’’ with the other person they were offered the chance ‘‘to work together and submit a proposal to the local movie theater’’ or ‘‘to go to a movie together as a group.’’ After receiving these instructions, participants viewed ten of the 15 total profiles created for the experiment. The decision to use only ten was made to reduce session length. Participants were then given a packet that contained a small version of each profile to use as a reference, a sheet asking for condition relevant desirability ratings for all ten of the people whose profiles they viewed, and 10
113
sheets that each contained a modified version of McCrosky and McCain’s (1974) measure of interpersonal attraction. Finally, they answered general demographic questions. After participants finished this, they were debriefed and thanked for participating. 2.4. Stimulus materials Fifteen avatars were selected for use in the study. Of the fifteen, three were nonhuman (one high on each of the three attractiveness cues: physical, social, and task), six were photographs (one male and one female high each attractiveness cue), and six were cartoonish humans (one male and one female high on each attractiveness cue) created through Yahoo! IM (please contact the first author to see each avatar). Selection began with the first and second authors each identifying multiple exemplars from public websites and by performing google searches for images (i.e. ‘‘attractive female photo’’) for each of the 15 avatar types needed for this study. The first and second authors then discussed their selections together, and decided what avatars should be used and which ones were still needed. This process continued until the 15 for use in the study were selected. The goal of this initial selection was to establish high face validity for the fit of the avatars to the experimental conditions varying gender, graphic form (nonhuman, and photograph and cartoon human) and task attractiveness cues (physical, social, and task). In order to increase realism, profiles were created to accompany the avatars used in this study. All profiles contained a brief description of a fictitious person and provided limited amounts of information about the purported interactant similar to the type of information one would typically see in social networking websites (e.g., name, gender, relationship status, personal interests, and an ‘About me. . .’-style biography entry). Ten intentionally vague descriptions (five male and five female) were created (see Fig. 1 for an example, see www.???????.edu for all fifteen avatars and all 10 profiles). Because the purpose of the avatar manipulation was to vary attractiveness based only on pictorial attributes of the avatar (uninfluenced by the text in the profile), seven different versions of the stimulus materials were created to control for unintended effects associated with specific descriptions. This was enough to provide a different set of avatar/profile combinations for each day of data collection and collect enough participants in each condition. Using stratified random techniques to match the gender of profile descriptor with visible attributes of avatar gender, the profiles accompanying the avatars were varied so that each profile co-occurred with the different avatars. 2.5. Measures Participants rated each person on how much they would like to choose that person for the specific goal (date, hang out, or work on a project) based on the 10 profiles they had just seen on a scale from 0 to 10. This rating was used as a measure of contextual-specific attraction. Thus, in the task condition, this rating was used as a measure of task attractiveness, in the social condition it was a measure of social attractiveness, and in the dating condition it was a measure of dating attractiveness. After viewing each profile, participants responded to a version of McCrosky and McCain’s (1974) measure of interpersonal attraction containing a response scale modified to provide a rating of uncertainty along each of three dimensions of attractiveness (physical, social, and task attractiveness) for each of the 10 ‘‘people’’ being evaluated. This measure was used for a couple of reasons. First, it is a general measure of attraction that taps into the three contexts of interest in the study. Second, it was utilized rather than the more recent McCroskey, McCroskey, and Richmond (2006) measure due to the length. The newer measure,
114
D. Westerman et al. / Computers in Human Behavior 53 (2015) 111–117
Fig. 1. Sample profile used in study.
although typically better, is also longer (38 vs. 18 items). Because participants would be responding to the measure multiple times in this study, concerns for burnout trumped the added quality of the more recent measure, and the McCroskey and McCain (1974) measure was used. The 18-item McCroskey and McCain (1974) measure includes five items to measure social attraction (i.e., I think he (she) could be a friend of mine), eight items to measure physical attraction (i.e., He (she) is very sexy looking), and five items to measure task attraction (I have confidence is his (her) ability to get the job done). The measure generally uses a five item response scale with 1 = strongly disagree and 5 = strongly agree. Borrowing a method of analysis that has been used in past impression formation research (Hancock & Dunham, 2001; Walther, 1993), the response scale was modified for the current study with the addition of a sixth response category, 6 = don’t know. The total number of ‘‘don’t know’’ responses was summed for each of the three types of attraction-uncertainty to create a separate uncertainty index for each type. The assumption behind this analysis is that respondents would be more likely to make judgments of any kind (as opposed to don’t know) if the avatar provides information reducing uncertainty for that judgment. Confirmatory factor analysis was conducted on each scale, and each of the three scales was unidimensional with acceptable reliability (social attractiveness, a = .80; physical attractiveness; a = .93; task attractiveness, a = .87) so all items were retained for analysis.
Two matching analyses also compared scores on social and physical attraction uncertainty for task attractive avatars. This same procedure was then repeated socially attractive avatars in the social context and physically attractive avatars in the dating context. To test H2’s prediction that attraction would increase when avatar cues conform to expectations created by an online-interaction context, three separate one-way ANOVAs were conducted on judgments of the likelihood the they would choose the person (represented by the avatar) as a partner for the interaction context given (dating attraction date, hang out, or work on a project). As such, within each interaction context (using scores from only respondents in that condition) analyses were conducted to examine the effect of conformity (as outlined in analyses for H1) on the likelihood of choice ratings of each profile. To address RQ1, separate one-way ANOVAs were conducted across avatar form (photographic human, cartoon human, nonhuman icon) using each of the three uncertainty measures (i.e., uncertainty regarding social, physical, and task attractiveness). To address RQ2, three separate one-way ANOVAs were conducted to examine the effect of avatar form on respondents’ judgments of the likelihood the they would choose the person (represented by the avatar) as a partner for the interaction context given (dating attraction date, hang out, or work on a project). As such, within each interaction context (and using scores from only respondents assigned to that context condition) analyses were conducted to examine the effect of graphic form (photographic human, cartoon human, nonhuman icon) on the likelihood of choice ratings of each profile.
2.6. Analysis plan 3. Results1 In order to test the hypotheses and address the research questions offered in the study, various ANOVA analyses were conducted. First, to test H1, nine separate one-way ANOVAs were conducted to examine the effect of avatar conformity on each of the three uncertainty variables (i.e., the social, physical and task attractiveness an avatar) within each of the three interaction contexts (dating attraction date, hang out, or work on a project). For example, in the task context, analyses compared scores on task uncertainty for avatars containing embedded task attractiveness cues with scores on task uncertainty for the avatars containing embedded social attractiveness or physical attractiveness cues.
3.1. The Effects of the graphic form of an avatar on uncertainty A significant effect for graphic form was found only for uncertainty about physical attractiveness F(2, 2048) = 329.42, p < .001, g2 = .24. Nonhuman icon avatars produced the highest uncertainty about physical attractiveness (M = 5.49, SD = 3.53), followed by 1 For the sake of clarity, only statistically significant findings are reported in text. For more complete data, including non-significant findings, please contact the first author.
D. Westerman et al. / Computers in Human Behavior 53 (2015) 111–117
cartoon human avatars (M = 3.93, SD = 3.76), and then photographic human avatars (M = 1.03, SD = 3.76.). 3.2. The effects of the graphic form of an avatar on attraction Results of these analyses show that graphic form had a significant difference on ratings of task attraction, F(2, 669) = 4.04, p < .05, g2 = .01. Nonhuman icon avatars had lower task attraction (M = 5.40, SD = 2.81), than both cartoon human avatars (M = 6.10, SD = 2.42), and photographic human avatars (M = 6.10, SD = 2.70). 3.3. The effects of avatar conformity with interaction context on uncertainty In the social condition, a significant difference was found for task attractiveness uncertainty, F(2, 709) = 3.49, p < .05, g2 = .01. Avatars designed for task attractiveness produced lower uncertainty about task attractiveness (M = 1.32, SD = 1.91) than avatars designed for physical attractiveness (M = 1.81, SD = 2.13). In the dating condition, once again, a significant difference was found for task attractiveness uncertainty, F(2, 675) = 5.34, p < .01, g2 = .02. Avatars designed for task attractiveness produced lower uncertainty about task attractiveness (M = 1.12, SD = 1.85) than avatars designed for both social attractiveness (M = 1.50, SD = 1.99) and physical attractiveness (M = 1.73, SD = 2.12). Overall, these data are not consistent with H1, as uncertainty does not differ in any condition in which the avatar cues were consistent to the context. However, the data do suggest that avatars designed for task attractiveness may help lower uncertainty about task attractiveness, as suggested by data in the social and dating context conditions. 3.4. The effects of avatar conformity with interaction context on attraction Results of these analyses show that avatar type had a significant effect on ratings of task attraction, F(2, 669) = 4.45, p < .05, g2 = .01. Avatars designed for social attractiveness had lower task attraction (M = 5.52, SD = 2.86), than both avatars designed for physical attractiveness (M = 6.13, SD = 2.39), and task attractiveness (M = 6.19, SD = 2.59). Significant differences were also found for avatar type on social attractiveness, F(2, 709) = 7.95, p < .001, g2 = .02. Avatars designed for both task attractiveness (M = 4.89, SD = 2.52) and social attractiveness (M = 5.35, SD = 2.58) had lower social attraction than avatars designed for physical attractiveness (M = 5.82, SD = 2.60). Although these data show that overall context matters for ratings of attractiveness, these differences are not in predicted manners. Thus, though the data suggest that avatar cues and context are important for attraction, these findings are not consistent with H2. 4. Discussion 4.1. Discussion of results Social information processing theory (Walther, 1992) states that people use whatever information they have available to them to help form judgments about others. However, the theory does not specify what information people use nor what role the context of impression formation plays in this process. The current research addressed these questions by looking at one source of visual information available in CMC: avatars. This is especially important because visual information is often heavily utilized in first impressions (Schneider et al., 1979) and online spaces such as SNS provide increased opportunity to choose how one wants to represent
115
oneself. The data provided in this study are consistent with the notion that both the context of impression formation and the graphic form of the avatar have effects on impression formation. More specifically, judgments of uncertainty and attraction were affected by both the graphic from of avatar and by the consistency between the context of impression formation and the attractiveness cues of the avatar, in line with past studies of judgments of avatar users (Donath, 2007; Nowak & Rauh, 2008; Weibel et al., 2010). Perhaps the strongest effect in this study was the effect that the graphic form of an avatar had on uncertainty about physical attractiveness. Across conditions, nonhuman avatars had the highest physical uncertainty, followed by cartoon humans, with photographs eliciting the lowest uncertainty about a supposed person’s physical attractiveness. These findings suggest one way that visual information may be used in forming impressions of others online: seeing is believing. Interestingly, cartoon humans provided a level of uncertainty that fell between photographs and icons, which suggests that some people were willing to make judgments about someone’s physical attractiveness based on a representation that is obviously not real, but is at least human or human-like. Human-like avatars have also shown an impact in research by Lortie and Guitton (2011, 2012), who found that they formed more homogeneous groups than other avatars in World of Warcraft. Perhaps if people are aware that a cartoon avatar is intended to be a physical representation of how a user believes they look, that person will use the information provided by the cartoon avatar to make judgments of the physical attraction of the avatar user, especially in the absence of photographs. Thus, the effect of these cartoon representations on ratings of physical attractiveness should be addressed in future research. Although strong differences were found for graphic form on uncertainty about physical attractiveness, they were not found for uncertainty about task or social attractiveness. This is interesting, as other research has suggested that anthropomorphism of an avatar impacts credibility of the user (Nowak & Rauh, 2008). One possible explanation for the lack of effects in the current study is that people may rely more on other information to make these judgments. For this study, participants were told that the profiles they saw were from ‘‘students in another class’’. This information may have been more heavily used by people to have confidence in their judgments about others’ social and task attractiveness. For example, knowing that a person was a student at the same university may have activated schemas about what a typical student is like there, which may have led to a belief that the person would be friendly and hard-working (or mean and lazy). Future research should examine what specific information is used to make these different contextual judgments. The graphic form of the avatar also had an effect on task attraction. Overall, iconic representations were rated as less task attractive than cartoon humans or photographs. Icons were likely seen as a less serious representation of a person than a human avatar (or even a cartoon human). Respondents were told that the fictitious profiles were created by people who were looking to chosen for a work project in the task attraction condition. It is possible that choosing an icon to represent oneself led respondents to question how serious that person was about working on the project, and thus led them to rate them lower in this regard. Future research can examine this reason for the finding. This study also examined how the consistency between impression formation context and attractiveness cues embedded in an avatar affected uncertainty and attraction. Although the data were not consistent with stated hypotheses, as uncertainty was not lower for any of the three conditions that exhibited an avatar attractiveness cues-context match, interesting differences were found for uncertainty about task attractiveness. Avatars created with task attractive cues embedded in them were rated as more
116
D. Westerman et al. / Computers in Human Behavior 53 (2015) 111–117
certain on task attractiveness than avatars designed for physical attractiveness (in both the social and dating condition) and social attractiveness (in the dating condition). Although this is interesting, it was also expected. More interesting is the fact that avatars designed to be both socially and physically attractive did not follow a similar pattern. Future research should investigate this anomaly. Conforming to expectations in an interaction context had an effect on ratings of attraction. For ratings of task attraction, avatars that were designed to be task or physically attractive were deemed higher than those that were created to be socially attractive. However, for ratings of social attraction, physically attractive avatars were rated higher than both task and socially attractive ones. These findings suggest a few things. First, it suggests that, similar to FtF interactions, a halo effect (Asch, 1946) might be operating among avatars. Avatars that were designed to be physically attractive were always more likely (or at least not less likely) to be selected for task, social or dating contexts, suggesting that people seemed to want to hang out with and work alongside others who had physically attractive avatars and want to work with them as well. That the halo effect manifests itself in online settings is consistent with Reeves and Nass’ (1996) notion of the media equation, which states that people respond to media like they do real life, and future research should examine this in greater detail. Analogously, these findings also highlight the need for continued studies examining multiple forms of information in impression formation. It is possible that manipulating the social attractiveness or task attractiveness of an avatar may increase the physical attractiveness of that avatar; conversely, manipulating the physical attractiveness of an avatar may in turn chance the task or social attractiveness of the same avatar. Visual characteristics of one’s avatar may also interact with other profile information, such as profiles statements posted by others (Walther et al., 2008) to change perceptions about a person. The notion of avatar-context match has relevance in other online spaces as well. For example, in World of Warcraft, Lortie and Guitton (2011, 2012) have found that players using more human-like avatars have a tendency to interact and affiliate with other human-like avatars, whereas less human-like avatars have a greater mix of characters in their groups and guilds. Players also tend to not leave homogenized groups. These are both of interest, as in terms of task-based performance in the game, heterogeneous groups seem more likely to lead to success. For example, there is a strong tendency toward homophily (e.g., McPherson, Smith-Lovin, & Cook, 2001) outside of MMORPGs as well, and this tendency seems to exist in World of Warcraft as well, also suggesting that is a space ripe for continued research in terms of avatars and context (social vs. task at least). Moving forward, the notion of the right avatar in the right context becomes challenging in some social networking sites. Whereas jeans and a T-shirt might be make a good first impression for a college student in a social setting, it is likely to be perceived as inappropriate business attire for somebody interviewing for an office job. Such a process should be replicated in a CMC environment, particularly in light of work suggesting that social networks often increased the amount of personal information that crosses across contexts as a function of increased SNS usage (context collapse: Vitak, 2012; Vitak, Lampe, Gray, & Ellison, 2012) – indeed, such work references the duality of wanting to be closer to one’s colleagues outside of the office by reducing interpersonal uncertainty (and using SNS to establish this social contact) while still balancing the risks of revealing much social information via SNS (to balance expectancy violations). How then might a potential employer conducting an online interview with someone for an office job respond to different type of avatars? If an interviewee selected an avatar in business attire, conformity with expectations created by the job interview context should reduce uncertainty, whereas the
presence of an avatar in jeans and a T-shirt should create doubt. Context collapse is a challenge for both researchers and users of these platforms, and needs to be considered moving forward. 4.2. Limitations As with all research the current study has some limitations, especially dealing with the generalizability of the sample. Although the current study was focused on relationships among variables, and is this likely more robust to sample issues than a study designed to measure the prevalence of variables in the population, the fact that the sample came from a communication course suggests that participants may not be naïve to theories and concepts underlying communication, both offline and online. This could impact the strength and directions of the relationships among variables found in the current study, and thus, future research can test the relationships found with different types of samples. A second potential issue with the sample is the female to male breakdown of the sample. Nearly 80% of the sample in the current study were female. A more equal distribution of males and females might lead to different results. This might be especially relevant in the dating condition. Many of the female avatars had few men rating them for their dating potential and physical attractiveness. A replication of this study with a more heterogeneous gender composition could be expected to change some of the physical attraction scores of the female avatars in this study. Also, this study did not specify whether or not the dating condition was intended to be short-term (one date) or long-term (a romantic partner). Research has found that women may choose men with different physical traits depending on the nature of their desired romantic relationship (Kruger, 2006). Future research should differentiate between these two types of dating relationships. 5. Conclusion Based on social information processing theory (Walther, 1992), which states that people use the cues that are available to them to form impressions of other people, this paper examined the effects of graphic form of avatars on impressions of uncertainty and attraction. Photographic avatars reduced more uncertainty about a person’s physical attractiveness than carton humans, which in turn reduced more uncertainty than icons. Icons were also rated as less task attractive than cartoon humans or photographs. It also examined the effects of impression formation context, and how a match between this context and the attractiveness cues of the avatar affect impression formation. Avatars designed to be task attractive had lower uncertainty about task attraction than those designed for physical attractiveness (in two conditions) and social attractiveness (in one condition). Data were also consistent with the idea that a halo effect (Asch, 1946) is occurring, as avatars embedded with physically attractive cues were rated high for both task and social attractiveness. Overall, the paper provides data that suggest that not only does avatar type have an impact on these judgments, but the context of the impression formation does as well. It suggests that not all information is responded to equally, and even the same information can have different effects based on other conditions, such as context. Future research extending SIPT should continue examining how information differentially impacts impression formation online. References Asch, S. E. (1946). Forming impressions of personality. Journal of Abnormal and Social Psychology, 41, 258–290. http://dx.doi.org/10.1037/h0055756. Bailenson, J., & Blascovich, J. (2004). Avatars. In W. S. Bainbridge (Ed.), Encyclopedia of human–computer interaction (pp. 64–68). Great Barrington, MA: Berkshire Publishing Group.
D. Westerman et al. / Computers in Human Behavior 53 (2015) 111–117 Berger, C. R., & Calabrese, R. J. (1975). Some explorations in initial interaction and beyond: Toward a developmental theory of interpersonal communication. Human Communication Research, 1, 99–112. http://dx.doi.org/10.1111/j.14682958.1975.tb00258.x. Berger, C. R. (1993). Uncertainty and social interaction. In S. A. Deetz (Ed.), Communication Yearbook 16 (pp. 491–502). Newbury Park, CA: Sage. Boyd, d. m., & Ellison, N. B. (2007). Social network sites: Definition, history, and scholarship. Journal of Computer-Mediated Communication, 13, 210–230. http:// dx.doi.org/10.1111/j.1083-6101.2007.00393.x. Chandler, D. (1998). Personal home pages and the construction of identities on the Web. Donath, J. (2007). Virtually trustworthy. Science, 317, 53–54. Facebook (2014). Company information. Facebook.com . Kotenko, J. (2013). Facebook reveals we upload a whopping 350 million photos to the network daily. DigitalTrends.com . Kruger, D. J. (2006). Male facial masculinity influences attributions of personality and reproductive strategy. Personal Relationships, 13, 451–463. http:// dx.doi.org/10.1111/j.1475-6811.2006.00129.x. Lee, E.-J. (2004). Effects of visual representation on social influence in computermediated communication: experimental tests of the social identity model of deindividuation effects. Human Communication Research, 30, 234–259. http:// dx.doi.org/10.1111/j.1468-2958.2004.tb00731.x. Lee, E.-J., & Nass, C. (2002). An experimental test of normative group influence and representation effects of computer-mediated communication: When interacting via computers differs from interaction with computers. Human Communication Research, 28, 349–381. http://dx.doi.org/10.1111/j.14682958.2002.tb00812.x. Lortie, C. L., & Guitton, M. J. (2011). Social organization in virtual settings depends on proximity to human visual aspect. Computers in Human Behavior, 27, 1258–1261. http://dx.doi.org/10.1016/j.chb.2011.01.006. Lortie, C. L., & Guitton, M. J. (2012). Looking similar promotes group stability in a game-based virtual community. Games for Health Journal: Research, Development, and Clinical Applications, 1, 274–278. http://dx.doi.org/10.1089/ g4h.2012.0016. Marcus, B., Machilek, F., & Schütz, A. (2006). Personality in cyberspace: Personal Web sites as media for personality expressions and impressions. Journal of Personality and Social Psychology, 90, 1014–1031. http://dx.doi.org/10.1037/ 0022-3514.90.6.1014. McCroskey, J. C., & McCain, T. A. (1974). The measurement of interpersonal attraction. Speech Monographs, 41, 261–266. http://dx.doi.org/10.1080/ 03637757409375845. McCroskey, L. L., McCroskey, J. C., & Richmond, V. P. (2006). Analysis and improvement of the measurement of interpersonal attraction and homophily. Communication Quarterly, 54, 1–31. http://dx.doi.org/10.1080/ 01463370500270322. McPherson, M., Smith-Lovin, L., & Cook, J. M. (2001). Birds of a feather: Homophily in social networks. Annual Review of Sociology, 27, 415–444. http://dx.doi.org/ 10.1146/annurev.soc.27.1.415.
117
Nowak, K. (2000). The influence of anthropomorphism on mental models of agents and avatars in social virtual environments. Unpublished doctoral dissertation. East Lansing, MI: Michigan State University. Nowak, K. L. (2004). The influence of anthropomorphism and agency on social judgment in virtual environments. Journal of Computer-Mediated Communication, 9(2), 00. http://dx.doi.org/10.1111/j.1083-6101.2004. tb00284.x. Nowak, K. L., & Rauh, C. (2005). The influence of the avatar on online perceptions of anthropomorphism, androgyny, credibility, homophily, and attraction. Journal of Computer-Mediated Communication, 11, 153–178. http://dx.doi.org/10.1111/ j.1083-6101.2006.tb00308.x. Nowak, K. L., & Rauh, C. (2008). Choose your ‘‘buddy icon’’: carefully: The influence of avatar androgyny, anthropomorphism and credibility in online interactions. Computers in Human Behavior, 24, 1473–1493. http://dx.doi.org/10.1016/ j.chb.2007.05.005. Planalp, S., & Honeycutt, J. (1985). Events that increase uncertainty in personal relationships. Human Communication Research, 11, 593–604. http://dx.doi.org/ 10.1111/j.1468-2958.1985.tb00062.x. Reeves, B., & Nass, C. (1996). The media equation: How people treat computers, television, and new media like real people and places. Stanford, CA: CSLI Publications. Rodriguez, S. (2013). Facebook reveals new News Feed design, emphasizes visual content. LA Times. Rubin, R. B. (1977). The role of context in information seeking and impression formation. Communication Monographs, 44, 81–90. http://dx.doi.org/10.1080/ 03637757709390117. Schneider, D. J., Hastorf, A. H., & Ellsworth, P. C. (1979). Person perception (2nd ed.). Menlo Park, CA: Addison-Wesley. Suler, J. (1997). The psychology of avatars and graphical space in multimedia chat communities . Tagiuri, R. (1969). Person perception. In G. L. Lindzay & E. Aronsen (Eds.), The handbook of social psychology (pp. 395–449). Reading, MA: Addison Wesley. Toma, C. L., Hancock, J. T., & Ellison, N. B. (2008). Separating fact from fiction: An examination of deceptive self-presentation in online dating profiles. Personality and Social Psychology Bulletin, 34, 1023–1036. http://dx.doi.org/10.1177/ 0146167208318067. Utz, S. (2010). Show me your friends and I will tell you what kind of person you are: How one’s profile, number of friends, and type of friends influence impression formation on social network sites. Journal of Computer-Mediated Communication, 15, 314–335. http://dx.doi.org/10.1111/j.1083-6101.2010.01522.x. Van Der Heide, B., D’Angelo, J. D., & Schumaker, E. M. (2012). The effects of verbal versus photographic self-presentation on impression formation in Facebook. Journal of Communication, 62, 98–116. http://dx.doi.org/10.1111/j.14602466.2011.01617.x. Vitak, J. (2012). The impact of context collapse and privacy on social network site disclosures. Journal of Broadcasting & Electronic Media, 56, 451–470. http:// dx.doi.org/10.1080/08838151.2012.732140. Vitak, J., Lampe, C., Gray, R., & Ellison, N. (2012). Why won’t you be my Facebook friend? Strategies for managing context collapse in the workplace. In Proceedings of conference ’12, Toronto, Canada . Walther, J. B. (1992). Interpersonal effects in computer-mediated interaction: A relational perspective. Communication Research, 19, 52–90. http://dx.doi.org/ 10.1177/009365092019001003. Walther, J. B. (1993). Construction and validation of a quantitative measure of impression development. Southern Communication Journal, 59, 27–33. http:// dx.doi.org/10.1080/10417949309372919. Walther, J. B. (1996). Computer-mediated communication: Impersonal, interpersonal, and hyperpersonal interaction. Communication Research, 23, 3–43. http://dx.doi.org/10.1177/009365096023001001. Walther, J. B., & Parks, M. R. (2002). Cues filtered out, cues filtered in: Computermediated communication and relationships. In M. L. Knapp & J. A. Daly (Eds.), Handbook of interpersonal communication (3rd ed., pp. 529–563). Thousand Oaks, CA: Sage. Walther, J. B., Van Der Heide, B., Kim, S.-Y., Westerman, D., & Tong, S. T. (2008). The role of friends’ behavior on evaluations of individuals’ Facebook profiles: Are we known by the company we keep? Human Communication Research, 34, 28–49. http://dx.doi.org/10.1111/j.1468-2958.2007.00312.x. Weibel, D., Stricker, D., Wissmath, B., & Mast, F. W. (2010). How socially relevant visual characteristics of avatars influence impression formation. Journal of Media Psychology, 22, 37–43. http://dx.doi.org/10.1027/1864-1105/a000005.
Contents lists available at ScienceDirect
Computers in Human Behavior journal homepage: www.elsevier.com/locate/comphumbeh
Full Length Article
The effects of static avatars on impression formation across different contexts on social networking sites David Westerman a,⇑, Ron Tamborini b, Nicholas David Bowman c a
North Dakota State University, Department of Communication, Department #2310, P.O. Box 6050, Fargo, ND 58108, United States Michigan State University, Department of Communication, 404 Wilson Road, Room 570, Communication Arts and Sciences Building, East Lansing, MI 48824, United States c West Virginia University, Department of Communication Studies, 108 Armstrong Hall, P.O. Box 6293, Morgantown, WV 26506, United States b
a r t i c l e
i n f o
Article history: Received 2 December 2014 Revised 10 June 2015 Accepted 12 June 2015 Available online 7 July 2015 Keywords: Avatars Uncertainty Attractiveness Impression formation
a b s t r a c t When making judgments about others, people use whatever social information is available in online environments. Such is the case for forming impressions of others. One type of such social information is a user’s avatar. This study examines different types of avatars (photographs, cartoon humans, and nonhumans) created for task, social or dating/romantic situations to study the effect of avatar type on judgments of uncertainty and task-specific attractiveness. Data suggest various patterns of uncertainty and attractiveness in these situations. Both the graphic from of an avatar and the context of impression formation have effects on subsequent impression formation. Judgments of uncertainty and attraction were affected by both the graphic from of avatar and by the consistency between the context of impression formation and the attractiveness cues of the avatar. These findings are discussed as are implications for future research. Ó 2015 Elsevier Ltd. All rights reserved.
1. Introduction
1.1. The use of visual information (avatars) in SNS
Social information processing theory (SIPT; Walther, 1992) assumes that people want to (and can) enact interpersonal processes online in order to attain interpersonal goals. One goal that is central to communicators – online or offline – is reducing uncertainty about others (Berger & Calabrese, 1975). In the text-based systems that typified CMC at the time of SIPT’s inception, people had few, if any, visual cues available to use for uncertainty reduction purposes. However, SIPT also suggests that people use the information that is available in an environment to accomplish their goals. Thus, even as more modern CMC systems provide increased visual cues – such as Facebook’s recent redesigns that emphasize visual information (Rodriguez, 2013), SIPT provides a useful framework to explain how users accomplish interpersonal goals online. The current study examined how increased visual information (of even a relatively low level) in a social networking site (SNS) can impact the impression formation process online.
Although some CMC is done through text-only channels (Walther & Parks, 2002), there are growing areas of CMC that incorporate visual information about communicators. One prominent online space utilizing such information is the social networking site (SNS). These sites provide cue-rich arenas for users to communicate using a mix of textual and visual cues. In fact, one of the hallmarks of such sites is that they allow people to ‘‘construct a public or semi-public profile’’ (boyd & Ellison, 2007, p. 211) and part of that profile construction is the inclusion of pictures. For example, Facebook, the largest SNS with over one billion monthly active users as of March 2014 (Facebook, 2014), recently reported that over 350 million unique photos are uploaded to its servers every day (Kotenko, 2013). One important piece of visual information that SNS users will provide is how they choose to represent themselves in their profile pictures. Visual appearance plays a big role in the impression formation process, both online and offline. This is especially true during first impressions of strangers, when nonverbal information can lead to spontaneous impressions of another person within a few seconds (Schneider, Hastorf, & Ellsworth, 1979) that can be very resistant to change (Kelley, 1950). Content analytic work by Hum et al. (2011) found that the majority of a sample of 150 college student Facebook profile pictures tended to be posed, inactive (not in motion), and containing only the subject. However, a profile
⇑ Corresponding author. E-mail addresses: [email protected] (D. Westerman), tamborin@ msu.edu (R. Tamborini), [email protected] (N.D. Bowman). http://dx.doi.org/10.1016/j.chb.2015.06.026 0747-5632/Ó 2015 Elsevier Ltd. All rights reserved.
112
D. Westerman et al. / Computers in Human Behavior 53 (2015) 111–117
picture does not have to be a photograph of the person. Instead, it can be what might be more generally referred to as an avatar. An avatar can be defined as a users’ graphical representations in a given virtual environment (Nowak, 2000), which can also include simple icons chosen as a form of self-representation (Suler, 1997). Overall, avatars can range from simple, static images to more animated and dynamic characters that are chosen to represent a person online (Bailenson & Blascovich, 2004). Avatars are different from agents in that avatars represent an actual person, whereas agents are computer-controlled entities. These various types of representations have a long history in online spaces, dating back at least to the use of icons, photographs, and other symbols in personal home pages (Chandler, 1998). Online spaces, such as SNS, provide users an increased opportunity to control their self-presentation (Walther, 1996). For example, people tend to exaggerate their attractiveness in their dating profiles, reporting the most deception in their profile pictures (Toma, Hancock, & Ellison, 2008). Donath (2007) also points out that people infer various character traits about other people based upon the visual appearance of an avatar. Van Der Heide, D’Angelo and Schumaker (2012) found that users privilege the photographic over textual information on Facebook profiles. Utz (2010) found that people were viewed as more socially attractive when their profile was more ‘‘extraverted’’, including a more extraverted picture. Friends’ profile pictures also had an impact on judgments of the profile owner (Utz, 2010; Walther, Van Der Heide, Kim, Westerman, & Tong, 2008). Notably, these are photographs used to represent people in various social profiles. What is unknown is how people will respond to various types of avatars that one might use to represent oneself in SNS. Thus, the following two research questions are offered: RQ1: How does the graphic form of an avatar (photograph, cartoon human, and nonhuman) influence a receiver’s uncertainty regarding the source? RQ2: How does the graphic form of an avatar (photograph, cartoon-human, and nonhuman) influence receiver perceptions of source attractiveness (social, physical, and task attractiveness)? 1.2. The role of context in judgments of people using avatars Past research shows that people make judgments based upon the avatars used in various platforms. These perceptions include intelligence (Koda, 2004), sociability and attractiveness (Weibel, Stricker, Wissmath, & Mast, 2010), personality traits (Marcus, Machilek, & Schütz, 2006), uncertainty (Nowak, 2004), credibility (Nowak & Rauh, 2005, 2008), group identity (Kim, 2009; Kim & Park, 2011; Lee; 2004; Lee & Nass, 2002) and affiliation (Lortie & Guitton, 2011; 2012). In general, the research suggests that people make judgments of avatars and the people that use them. Although people may make judgments of others based on visual cues, this is not the only source of information that may be used. Tagiuri (1969) suggests that people form impressions of others by a combination of object cues and situational context. In fact, past studies involving avatars in CMC (Nowak & Rauh, 2005) have called for more research examining the effects of different contexts on impression formation. We might expect that avatars created to provide information that was consistent with goals of the interaction context would do a better job of reducing uncertainty and increasing liking. Why might this be the case? Generally stated, information that conforms to expectations should reduce uncertainty, whereas an expectancy violation may increase it. Expectancy violations tend to increase uncertainty (Berger, 1993; Planalp & Honeycutt, 1985). Applied to CMC, an avatar that fits preconceived notions
of an interaction context should reduce uncertainty, whereas one that violates preconceived notions of what to expect in a given context may raise questions and increase uncertainty. In FtF settings, an ambiguous interaction context has been found to create more uncertainty than a more specific, task-based context (Rubin, 1977). This suggests that context of an interaction can reduce uncertainty by providing a focal point for initial interaction. In other terms, context helps specify which information in an interaction is important to a task (Kelly, 1955). Without an understood context for interaction, it is difficult to imagine what information participants in previous research might have gained from avatars to help reduce uncertainty or form initial impressions about their interactional partner. At a broader level, the basic information one usually provides on a social network – starting with the earliest MySpace pages to more popular Facebook profiles – may be thought of in terms of McCroskey and McCain’s (1974) classic measure of interpersonal attraction: namely, social attraction, physical attraction, and task attraction. Most SNSs, even sites more aligned with professional rather than personal networking such as LinkedIN, request one to provide their physical picture (physical attractiveness) as well as their hobbies and interests (social attractiveness) as well as their education and work experience (task attractiveness). The focus on these features of social interaction points to their importance in interpersonal communication and calls our attention to them in efforts to understand online communication. We presume that different expectations for an interaction are established by the belief that an online exchange is motivated by one purpose versus another. Moreover, we should anticipate that uncertainty reduction is governed by the extent to which information received during an online exchange conforms to expectations created by these different contexts. Information that conforms to expectations should reduce uncertainty, whereas an expectancy violation may increase it. In this manner, information interacts with context to influence uncertainty reduction. For example, cues that conform to expectations for a social interaction task might conflict with expectations for a business related task. This leads to the following hypotheses: H1. The context of an online interaction (potential dating, hanging out, or task achievement) interacts with cues present in an online avatar (physical, social, and task attractiveness) to influence perceptions of uncertainty regarding the source such that cues that conflict with expectations for the interaction context lead to greater uncertainty regarding the source than cues that conform to those expectations.
H2. The context of an online interaction (potential dating, hanging out, or task achievement) interacts with cues present in an online avatar (physical, social, and task attractiveness) to influence perceptions of source attractiveness such that cues that conform to expectations for the interaction context lead to greater source attractiveness than cues that conflict with those expectations.
2. Method 2.1. Overview A 3 3 mixed design varied the context given to respondents for online interaction and the attractiveness cues of avatars attributed to 15 apparent online partners. The between subjects factor was context. Respondents were randomly assigned to one of three conditions varying the reason given for their online interaction (for potential dating, hanging out, or task achievement). The within
D. Westerman et al. / Computers in Human Behavior 53 (2015) 111–117
subject factor varied three dimensions of attractiveness (physical, social, and task attractiveness) in the avatars by manipulating each avatar related to physical good looks, social gregariousness, and task achievement. Respondents were asked to evaluate those individuals online who were purported to have chosen each of the 15 different avatars using scales designed to measure three dimensions of uncertainty and attractiveness. 2.2. Participants 206 undergraduates enrolled in introductory communication courses at a large Midwestern university participated in this study in exchange for course credit. The sample ranged in age from 17 to 24 (M = 19.95, SD = 1.30). Most of the sample (80.1%) was female with three people (1.5%) not responding. Racially, the sample was comprised of 22 (10.7%) African-Americans, five (2.4%) Asian-Americans, 164 (79.6%) Caucasians, four (1.9%) Hispanics, three (1.5%) mixed, and seven (3.4%) others; one person (0.5%) did not respond. 2.3. Procedure Participants were told that they would be offering comments on profiles created by students ‘‘in another class.’’ They were seated together in groups ranging in size from seven to 13, but did not interact with each other. After assignment to one of the three interaction settings, respondents were given instructions that primed attention to the profile creators’ physical, social, or task attractiveness. The instructions for the dating condition (physical attractiveness) were as follows: ‘‘Thank you for your willingness to participate in this study. Today, you are going to be asked to participate in a study similar to speed dating, only online. Students in another class were asked to create profiles for this speed dating. You are about to see ten of these profiles. After you have viewed all 10 profiles, you will be asked to rate each person for how much you would potentially like to go on a date with them. Please note that you are being randomly assigned profiles from this class, so you may see profiles of both males and females. After you rate each person, you will be asked to answer some questions about each one, and then some questions about yourself. At the end of the entire study, four couples will be randomly chosen and will be offered the chance to go on a movie date. Those couples will be chosen by taking the people with the four highest rated profiles, and randomly selecting one person that had them ranked as their top choice.’’ The instructions for the task achievement condition (task attractiveness) and the hanging-out condition (social attractiveness) were similar to those for the dating condition with the following exceptions: First, instead of respondents being asked to ‘‘participate in a study similar to speed dating,’’ in the task condition they were asked to ‘‘participate in a study that may give you a chance to work with a local movie theater in promoting a local film that will be playing in January,’’ and in the hanging-out condition they were asked to ‘‘choose a group of people you think would be fun to hang out with.’’ Also, instead of being asked to rate how much they would like ‘‘to go on a date with’’ the other person they were asked to rate how much they would like ‘‘to work with them’’ or ‘‘to hang out with them.’’ Finally, instead of being offered the chance ‘‘to go on a movie date’’ with the other person they were offered the chance ‘‘to work together and submit a proposal to the local movie theater’’ or ‘‘to go to a movie together as a group.’’ After receiving these instructions, participants viewed ten of the 15 total profiles created for the experiment. The decision to use only ten was made to reduce session length. Participants were then given a packet that contained a small version of each profile to use as a reference, a sheet asking for condition relevant desirability ratings for all ten of the people whose profiles they viewed, and 10
113
sheets that each contained a modified version of McCrosky and McCain’s (1974) measure of interpersonal attraction. Finally, they answered general demographic questions. After participants finished this, they were debriefed and thanked for participating. 2.4. Stimulus materials Fifteen avatars were selected for use in the study. Of the fifteen, three were nonhuman (one high on each of the three attractiveness cues: physical, social, and task), six were photographs (one male and one female high each attractiveness cue), and six were cartoonish humans (one male and one female high on each attractiveness cue) created through Yahoo! IM (please contact the first author to see each avatar). Selection began with the first and second authors each identifying multiple exemplars from public websites and by performing google searches for images (i.e. ‘‘attractive female photo’’) for each of the 15 avatar types needed for this study. The first and second authors then discussed their selections together, and decided what avatars should be used and which ones were still needed. This process continued until the 15 for use in the study were selected. The goal of this initial selection was to establish high face validity for the fit of the avatars to the experimental conditions varying gender, graphic form (nonhuman, and photograph and cartoon human) and task attractiveness cues (physical, social, and task). In order to increase realism, profiles were created to accompany the avatars used in this study. All profiles contained a brief description of a fictitious person and provided limited amounts of information about the purported interactant similar to the type of information one would typically see in social networking websites (e.g., name, gender, relationship status, personal interests, and an ‘About me. . .’-style biography entry). Ten intentionally vague descriptions (five male and five female) were created (see Fig. 1 for an example, see www.???????.edu for all fifteen avatars and all 10 profiles). Because the purpose of the avatar manipulation was to vary attractiveness based only on pictorial attributes of the avatar (uninfluenced by the text in the profile), seven different versions of the stimulus materials were created to control for unintended effects associated with specific descriptions. This was enough to provide a different set of avatar/profile combinations for each day of data collection and collect enough participants in each condition. Using stratified random techniques to match the gender of profile descriptor with visible attributes of avatar gender, the profiles accompanying the avatars were varied so that each profile co-occurred with the different avatars. 2.5. Measures Participants rated each person on how much they would like to choose that person for the specific goal (date, hang out, or work on a project) based on the 10 profiles they had just seen on a scale from 0 to 10. This rating was used as a measure of contextual-specific attraction. Thus, in the task condition, this rating was used as a measure of task attractiveness, in the social condition it was a measure of social attractiveness, and in the dating condition it was a measure of dating attractiveness. After viewing each profile, participants responded to a version of McCrosky and McCain’s (1974) measure of interpersonal attraction containing a response scale modified to provide a rating of uncertainty along each of three dimensions of attractiveness (physical, social, and task attractiveness) for each of the 10 ‘‘people’’ being evaluated. This measure was used for a couple of reasons. First, it is a general measure of attraction that taps into the three contexts of interest in the study. Second, it was utilized rather than the more recent McCroskey, McCroskey, and Richmond (2006) measure due to the length. The newer measure,
114
D. Westerman et al. / Computers in Human Behavior 53 (2015) 111–117
Fig. 1. Sample profile used in study.
although typically better, is also longer (38 vs. 18 items). Because participants would be responding to the measure multiple times in this study, concerns for burnout trumped the added quality of the more recent measure, and the McCroskey and McCain (1974) measure was used. The 18-item McCroskey and McCain (1974) measure includes five items to measure social attraction (i.e., I think he (she) could be a friend of mine), eight items to measure physical attraction (i.e., He (she) is very sexy looking), and five items to measure task attraction (I have confidence is his (her) ability to get the job done). The measure generally uses a five item response scale with 1 = strongly disagree and 5 = strongly agree. Borrowing a method of analysis that has been used in past impression formation research (Hancock & Dunham, 2001; Walther, 1993), the response scale was modified for the current study with the addition of a sixth response category, 6 = don’t know. The total number of ‘‘don’t know’’ responses was summed for each of the three types of attraction-uncertainty to create a separate uncertainty index for each type. The assumption behind this analysis is that respondents would be more likely to make judgments of any kind (as opposed to don’t know) if the avatar provides information reducing uncertainty for that judgment. Confirmatory factor analysis was conducted on each scale, and each of the three scales was unidimensional with acceptable reliability (social attractiveness, a = .80; physical attractiveness; a = .93; task attractiveness, a = .87) so all items were retained for analysis.
Two matching analyses also compared scores on social and physical attraction uncertainty for task attractive avatars. This same procedure was then repeated socially attractive avatars in the social context and physically attractive avatars in the dating context. To test H2’s prediction that attraction would increase when avatar cues conform to expectations created by an online-interaction context, three separate one-way ANOVAs were conducted on judgments of the likelihood the they would choose the person (represented by the avatar) as a partner for the interaction context given (dating attraction date, hang out, or work on a project). As such, within each interaction context (using scores from only respondents in that condition) analyses were conducted to examine the effect of conformity (as outlined in analyses for H1) on the likelihood of choice ratings of each profile. To address RQ1, separate one-way ANOVAs were conducted across avatar form (photographic human, cartoon human, nonhuman icon) using each of the three uncertainty measures (i.e., uncertainty regarding social, physical, and task attractiveness). To address RQ2, three separate one-way ANOVAs were conducted to examine the effect of avatar form on respondents’ judgments of the likelihood the they would choose the person (represented by the avatar) as a partner for the interaction context given (dating attraction date, hang out, or work on a project). As such, within each interaction context (and using scores from only respondents assigned to that context condition) analyses were conducted to examine the effect of graphic form (photographic human, cartoon human, nonhuman icon) on the likelihood of choice ratings of each profile.
2.6. Analysis plan 3. Results1 In order to test the hypotheses and address the research questions offered in the study, various ANOVA analyses were conducted. First, to test H1, nine separate one-way ANOVAs were conducted to examine the effect of avatar conformity on each of the three uncertainty variables (i.e., the social, physical and task attractiveness an avatar) within each of the three interaction contexts (dating attraction date, hang out, or work on a project). For example, in the task context, analyses compared scores on task uncertainty for avatars containing embedded task attractiveness cues with scores on task uncertainty for the avatars containing embedded social attractiveness or physical attractiveness cues.
3.1. The Effects of the graphic form of an avatar on uncertainty A significant effect for graphic form was found only for uncertainty about physical attractiveness F(2, 2048) = 329.42, p < .001, g2 = .24. Nonhuman icon avatars produced the highest uncertainty about physical attractiveness (M = 5.49, SD = 3.53), followed by 1 For the sake of clarity, only statistically significant findings are reported in text. For more complete data, including non-significant findings, please contact the first author.
D. Westerman et al. / Computers in Human Behavior 53 (2015) 111–117
cartoon human avatars (M = 3.93, SD = 3.76), and then photographic human avatars (M = 1.03, SD = 3.76.). 3.2. The effects of the graphic form of an avatar on attraction Results of these analyses show that graphic form had a significant difference on ratings of task attraction, F(2, 669) = 4.04, p < .05, g2 = .01. Nonhuman icon avatars had lower task attraction (M = 5.40, SD = 2.81), than both cartoon human avatars (M = 6.10, SD = 2.42), and photographic human avatars (M = 6.10, SD = 2.70). 3.3. The effects of avatar conformity with interaction context on uncertainty In the social condition, a significant difference was found for task attractiveness uncertainty, F(2, 709) = 3.49, p < .05, g2 = .01. Avatars designed for task attractiveness produced lower uncertainty about task attractiveness (M = 1.32, SD = 1.91) than avatars designed for physical attractiveness (M = 1.81, SD = 2.13). In the dating condition, once again, a significant difference was found for task attractiveness uncertainty, F(2, 675) = 5.34, p < .01, g2 = .02. Avatars designed for task attractiveness produced lower uncertainty about task attractiveness (M = 1.12, SD = 1.85) than avatars designed for both social attractiveness (M = 1.50, SD = 1.99) and physical attractiveness (M = 1.73, SD = 2.12). Overall, these data are not consistent with H1, as uncertainty does not differ in any condition in which the avatar cues were consistent to the context. However, the data do suggest that avatars designed for task attractiveness may help lower uncertainty about task attractiveness, as suggested by data in the social and dating context conditions. 3.4. The effects of avatar conformity with interaction context on attraction Results of these analyses show that avatar type had a significant effect on ratings of task attraction, F(2, 669) = 4.45, p < .05, g2 = .01. Avatars designed for social attractiveness had lower task attraction (M = 5.52, SD = 2.86), than both avatars designed for physical attractiveness (M = 6.13, SD = 2.39), and task attractiveness (M = 6.19, SD = 2.59). Significant differences were also found for avatar type on social attractiveness, F(2, 709) = 7.95, p < .001, g2 = .02. Avatars designed for both task attractiveness (M = 4.89, SD = 2.52) and social attractiveness (M = 5.35, SD = 2.58) had lower social attraction than avatars designed for physical attractiveness (M = 5.82, SD = 2.60). Although these data show that overall context matters for ratings of attractiveness, these differences are not in predicted manners. Thus, though the data suggest that avatar cues and context are important for attraction, these findings are not consistent with H2. 4. Discussion 4.1. Discussion of results Social information processing theory (Walther, 1992) states that people use whatever information they have available to them to help form judgments about others. However, the theory does not specify what information people use nor what role the context of impression formation plays in this process. The current research addressed these questions by looking at one source of visual information available in CMC: avatars. This is especially important because visual information is often heavily utilized in first impressions (Schneider et al., 1979) and online spaces such as SNS provide increased opportunity to choose how one wants to represent
115
oneself. The data provided in this study are consistent with the notion that both the context of impression formation and the graphic form of the avatar have effects on impression formation. More specifically, judgments of uncertainty and attraction were affected by both the graphic from of avatar and by the consistency between the context of impression formation and the attractiveness cues of the avatar, in line with past studies of judgments of avatar users (Donath, 2007; Nowak & Rauh, 2008; Weibel et al., 2010). Perhaps the strongest effect in this study was the effect that the graphic form of an avatar had on uncertainty about physical attractiveness. Across conditions, nonhuman avatars had the highest physical uncertainty, followed by cartoon humans, with photographs eliciting the lowest uncertainty about a supposed person’s physical attractiveness. These findings suggest one way that visual information may be used in forming impressions of others online: seeing is believing. Interestingly, cartoon humans provided a level of uncertainty that fell between photographs and icons, which suggests that some people were willing to make judgments about someone’s physical attractiveness based on a representation that is obviously not real, but is at least human or human-like. Human-like avatars have also shown an impact in research by Lortie and Guitton (2011, 2012), who found that they formed more homogeneous groups than other avatars in World of Warcraft. Perhaps if people are aware that a cartoon avatar is intended to be a physical representation of how a user believes they look, that person will use the information provided by the cartoon avatar to make judgments of the physical attraction of the avatar user, especially in the absence of photographs. Thus, the effect of these cartoon representations on ratings of physical attractiveness should be addressed in future research. Although strong differences were found for graphic form on uncertainty about physical attractiveness, they were not found for uncertainty about task or social attractiveness. This is interesting, as other research has suggested that anthropomorphism of an avatar impacts credibility of the user (Nowak & Rauh, 2008). One possible explanation for the lack of effects in the current study is that people may rely more on other information to make these judgments. For this study, participants were told that the profiles they saw were from ‘‘students in another class’’. This information may have been more heavily used by people to have confidence in their judgments about others’ social and task attractiveness. For example, knowing that a person was a student at the same university may have activated schemas about what a typical student is like there, which may have led to a belief that the person would be friendly and hard-working (or mean and lazy). Future research should examine what specific information is used to make these different contextual judgments. The graphic form of the avatar also had an effect on task attraction. Overall, iconic representations were rated as less task attractive than cartoon humans or photographs. Icons were likely seen as a less serious representation of a person than a human avatar (or even a cartoon human). Respondents were told that the fictitious profiles were created by people who were looking to chosen for a work project in the task attraction condition. It is possible that choosing an icon to represent oneself led respondents to question how serious that person was about working on the project, and thus led them to rate them lower in this regard. Future research can examine this reason for the finding. This study also examined how the consistency between impression formation context and attractiveness cues embedded in an avatar affected uncertainty and attraction. Although the data were not consistent with stated hypotheses, as uncertainty was not lower for any of the three conditions that exhibited an avatar attractiveness cues-context match, interesting differences were found for uncertainty about task attractiveness. Avatars created with task attractive cues embedded in them were rated as more
116
D. Westerman et al. / Computers in Human Behavior 53 (2015) 111–117
certain on task attractiveness than avatars designed for physical attractiveness (in both the social and dating condition) and social attractiveness (in the dating condition). Although this is interesting, it was also expected. More interesting is the fact that avatars designed to be both socially and physically attractive did not follow a similar pattern. Future research should investigate this anomaly. Conforming to expectations in an interaction context had an effect on ratings of attraction. For ratings of task attraction, avatars that were designed to be task or physically attractive were deemed higher than those that were created to be socially attractive. However, for ratings of social attraction, physically attractive avatars were rated higher than both task and socially attractive ones. These findings suggest a few things. First, it suggests that, similar to FtF interactions, a halo effect (Asch, 1946) might be operating among avatars. Avatars that were designed to be physically attractive were always more likely (or at least not less likely) to be selected for task, social or dating contexts, suggesting that people seemed to want to hang out with and work alongside others who had physically attractive avatars and want to work with them as well. That the halo effect manifests itself in online settings is consistent with Reeves and Nass’ (1996) notion of the media equation, which states that people respond to media like they do real life, and future research should examine this in greater detail. Analogously, these findings also highlight the need for continued studies examining multiple forms of information in impression formation. It is possible that manipulating the social attractiveness or task attractiveness of an avatar may increase the physical attractiveness of that avatar; conversely, manipulating the physical attractiveness of an avatar may in turn chance the task or social attractiveness of the same avatar. Visual characteristics of one’s avatar may also interact with other profile information, such as profiles statements posted by others (Walther et al., 2008) to change perceptions about a person. The notion of avatar-context match has relevance in other online spaces as well. For example, in World of Warcraft, Lortie and Guitton (2011, 2012) have found that players using more human-like avatars have a tendency to interact and affiliate with other human-like avatars, whereas less human-like avatars have a greater mix of characters in their groups and guilds. Players also tend to not leave homogenized groups. These are both of interest, as in terms of task-based performance in the game, heterogeneous groups seem more likely to lead to success. For example, there is a strong tendency toward homophily (e.g., McPherson, Smith-Lovin, & Cook, 2001) outside of MMORPGs as well, and this tendency seems to exist in World of Warcraft as well, also suggesting that is a space ripe for continued research in terms of avatars and context (social vs. task at least). Moving forward, the notion of the right avatar in the right context becomes challenging in some social networking sites. Whereas jeans and a T-shirt might be make a good first impression for a college student in a social setting, it is likely to be perceived as inappropriate business attire for somebody interviewing for an office job. Such a process should be replicated in a CMC environment, particularly in light of work suggesting that social networks often increased the amount of personal information that crosses across contexts as a function of increased SNS usage (context collapse: Vitak, 2012; Vitak, Lampe, Gray, & Ellison, 2012) – indeed, such work references the duality of wanting to be closer to one’s colleagues outside of the office by reducing interpersonal uncertainty (and using SNS to establish this social contact) while still balancing the risks of revealing much social information via SNS (to balance expectancy violations). How then might a potential employer conducting an online interview with someone for an office job respond to different type of avatars? If an interviewee selected an avatar in business attire, conformity with expectations created by the job interview context should reduce uncertainty, whereas the
presence of an avatar in jeans and a T-shirt should create doubt. Context collapse is a challenge for both researchers and users of these platforms, and needs to be considered moving forward. 4.2. Limitations As with all research the current study has some limitations, especially dealing with the generalizability of the sample. Although the current study was focused on relationships among variables, and is this likely more robust to sample issues than a study designed to measure the prevalence of variables in the population, the fact that the sample came from a communication course suggests that participants may not be naïve to theories and concepts underlying communication, both offline and online. This could impact the strength and directions of the relationships among variables found in the current study, and thus, future research can test the relationships found with different types of samples. A second potential issue with the sample is the female to male breakdown of the sample. Nearly 80% of the sample in the current study were female. A more equal distribution of males and females might lead to different results. This might be especially relevant in the dating condition. Many of the female avatars had few men rating them for their dating potential and physical attractiveness. A replication of this study with a more heterogeneous gender composition could be expected to change some of the physical attraction scores of the female avatars in this study. Also, this study did not specify whether or not the dating condition was intended to be short-term (one date) or long-term (a romantic partner). Research has found that women may choose men with different physical traits depending on the nature of their desired romantic relationship (Kruger, 2006). Future research should differentiate between these two types of dating relationships. 5. Conclusion Based on social information processing theory (Walther, 1992), which states that people use the cues that are available to them to form impressions of other people, this paper examined the effects of graphic form of avatars on impressions of uncertainty and attraction. Photographic avatars reduced more uncertainty about a person’s physical attractiveness than carton humans, which in turn reduced more uncertainty than icons. Icons were also rated as less task attractive than cartoon humans or photographs. It also examined the effects of impression formation context, and how a match between this context and the attractiveness cues of the avatar affect impression formation. Avatars designed to be task attractive had lower uncertainty about task attraction than those designed for physical attractiveness (in two conditions) and social attractiveness (in one condition). Data were also consistent with the idea that a halo effect (Asch, 1946) is occurring, as avatars embedded with physically attractive cues were rated high for both task and social attractiveness. Overall, the paper provides data that suggest that not only does avatar type have an impact on these judgments, but the context of the impression formation does as well. It suggests that not all information is responded to equally, and even the same information can have different effects based on other conditions, such as context. Future research extending SIPT should continue examining how information differentially impacts impression formation online. References Asch, S. E. (1946). Forming impressions of personality. Journal of Abnormal and Social Psychology, 41, 258–290. http://dx.doi.org/10.1037/h0055756. Bailenson, J., & Blascovich, J. (2004). Avatars. In W. S. Bainbridge (Ed.), Encyclopedia of human–computer interaction (pp. 64–68). Great Barrington, MA: Berkshire Publishing Group.
D. Westerman et al. / Computers in Human Behavior 53 (2015) 111–117 Berger, C. R., & Calabrese, R. J. (1975). Some explorations in initial interaction and beyond: Toward a developmental theory of interpersonal communication. Human Communication Research, 1, 99–112. http://dx.doi.org/10.1111/j.14682958.1975.tb00258.x. Berger, C. R. (1993). Uncertainty and social interaction. In S. A. Deetz (Ed.), Communication Yearbook 16 (pp. 491–502). Newbury Park, CA: Sage. Boyd, d. m., & Ellison, N. B. (2007). Social network sites: Definition, history, and scholarship. Journal of Computer-Mediated Communication, 13, 210–230. http:// dx.doi.org/10.1111/j.1083-6101.2007.00393.x. Chandler, D. (1998). Personal home pages and the construction of identities on the Web
117
Nowak, K. (2000). The influence of anthropomorphism on mental models of agents and avatars in social virtual environments. Unpublished doctoral dissertation. East Lansing, MI: Michigan State University. Nowak, K. L. (2004). The influence of anthropomorphism and agency on social judgment in virtual environments. Journal of Computer-Mediated Communication, 9(2), 00. http://dx.doi.org/10.1111/j.1083-6101.2004. tb00284.x. Nowak, K. L., & Rauh, C. (2005). The influence of the avatar on online perceptions of anthropomorphism, androgyny, credibility, homophily, and attraction. Journal of Computer-Mediated Communication, 11, 153–178. http://dx.doi.org/10.1111/ j.1083-6101.2006.tb00308.x. Nowak, K. L., & Rauh, C. (2008). Choose your ‘‘buddy icon’’: carefully: The influence of avatar androgyny, anthropomorphism and credibility in online interactions. Computers in Human Behavior, 24, 1473–1493. http://dx.doi.org/10.1016/ j.chb.2007.05.005. Planalp, S., & Honeycutt, J. (1985). Events that increase uncertainty in personal relationships. Human Communication Research, 11, 593–604. http://dx.doi.org/ 10.1111/j.1468-2958.1985.tb00062.x. Reeves, B., & Nass, C. (1996). The media equation: How people treat computers, television, and new media like real people and places. Stanford, CA: CSLI Publications. Rodriguez, S. (2013). Facebook reveals new News Feed design, emphasizes visual content. LA Times