avatars in social media

Int. J. Human-Computer Studies 66 (2008) 801811

Avatars in social media: Balancing accuracy, playfulness andembodied messages

Asimina Vasaloua,!, Adam Joinsona, Tanja Banzigerb, Peter Goldiec, Jeremy Pittd

aSchool of Information Management, University of Bath, BA2 7AA, Bath, UKbDepartment of Psychology, University of Geneva, 40 Boulevard du Pont-dArve, CH-1205, Geneva, Switzerland

cDepartment of Philosophy, University of Manchester, Dover Street Building, M13 9PL, Manchester, UKdElectrical and Electronic Engineering, Imperial College, Exhibition Road, SW7 2AZ, London, UK

Received 9 May 2008; received in revised form 6 August 2008; accepted 11 August 2008Communicated by S. WiedenbeckAvailable online 14 August 2008

Abstract

This paper examines how users negotiate their self-presentation via an avatar used in social media. Twenty participants customised anavatar while thinking aloud. An analysis of this verbal data revealed three motivating factors that drive self-presentation: (1) avatars wereused to accurately reflect their owners offline self; participants chose to display stable self-attributes or idealised their avatar byconcealing or emphasising attributes aligned to imagined social roles, (2) the diversity of customisation options was exploited by someparticipants who broke free from the social rules governing self-presentation offline; others used the avatars appearance to emotionallyprovoke and engage the avatar viewer and finally, (3) avatars were used as proxies; participants designed their online self in order toconvey a message to a significant other.r 2008 Elsevier Ltd. All rights reserved.

Keywords: Avatars; Social media; Self-presentation; Identity

1. Introduction

Increasingly, social system designers are introducingavatars as the new forum for identity construction in onlineenvironments, a feature users have welcomed with en-thusiasm. Following Yahoos launch of Yahoo! Avatars, 7million users personalised an avatar for display within theirprofiles e.g. Yahoo! 3601, Yahoo! Answers, Yahoo!Messenger (Hemp, 2006). Likewise, four million userscreated an avatar after the facility (called weemee) wasintroduced to the chat system Windows Live Messenger.1

This rapid uptake of avatars is not surprising; while avatarsmaintain users privacy, they give expressive freedom over

an otherwise static online identity. In an illustrativeexample, V-Chat, a virtual chat-room, allowed users toeither design their own avatar, which was particularly timedemanding, or to select from already-made options. Ingeneral, V-Chat users spurned the latter option, insteadtaking the time to create unique representations so thatthey didnt look like others (Cheng et al., 2002).Avatars are malleable objects as users are given the

ability to choose diverse customisation combinations.Motivated by this fluid property of avatars, HCI research-ers have examined the social implications of avatars byconsidering how users manage self-presentation throughthis new medium and to what extent an avatarsappearance can shape online communication (e.g. Bers etal., 2001; Axelsson, 2002; Taylor, 2002; Nowak and Rauh,2005). This prior work has traditionally taken place invirtual environments where users avatars serve as onlineembodiments for navigating three-dimensional spaces.Though members may forge friendships over time, at theonset they communicate with previously unknown others.

ARTICLE IN PRESS

www.elsevier.com/locate/ijhcs

1071-5819/$ - see front matter r 2008 Elsevier Ltd. All rights reserved.doi:10.1016/j.ijhcs.2008.08.002

!Corresponding author.E-mail addresses: [email protected] (A. Vasalou), a.joinson@

bath.ac.uk (A. Joinson), [email protected] (T. Banziger),[email protected] (P. Goldie), [email protected](J. Pitt).

1Retrieved from http://www.weeworld.com/about/partner.aspx?partner 22 on 31/03/08.

More recently, avatars have been implanted in social mediaenvironments. Unlike the strangers one encounters in otheronline environments, users of social media predominantlyinteract with offline contacts such as friends and family:bloggers document their daily lives to keep their family andfriends informed (Nardi et al., 2004); users of the socialnetworking site Facebook visit the site to reconnect withold friends or to maintain ongoing friendships (Joinson,2008). Furthermore, the open text channel available insocial media provokes self-reflection that is reminiscentof diary entries: bloggers document their life events (Nardiet al., 2004), use blogging as a form of catharsis forworking out their emotions (Nardi et al., 2004; Huffakerand Calvert, 2005), or voice their opinions on controversialtopics e.g. politics (Nardi et al., 2004). Therefore, in socialmedia, ones offline social and emotional life is merged withthe online sphere.This unity of offline and online has been found to

facilitate more honest constructions of identity and self-presentation, as opposed to the role playing that oftentakes place in virtual environments or MUDs (e.g.Bruckman, 1993; Turkle, 1995; Taylor, 2002).2 To givean example, bloggers frequently report identifying andaccurate profile information such as their real name or age(Herring et al., 2004; Huffaker and Calvert, 2005).A similar result has been shown in relation to avatars;users tend to reflect their own appearance when personalis-ing an avatar for display within their blog (Vasalou et al.,2007). This previous work, however, does not reveal thestrategies that users draw together to shape a self-representative avatar. The present research addresses thispoint with a qualitative study that captures the array ofself-presentation strategies that users apply when con-structing an avatar. By elucidating the how behindself-presentation, we gain a more profound theoreticalunderstanding about identity in social media. At the sametime, by exposing which customisation choices are sig-nificant to users, we extrapolate design recommendationsfor avatar customisation tools.We begin by discussing how theories on identity and self-

presentation apply to the unique qualities of the avatarmedium. Next, we describe the experiment conducted;users were requested to construct an avatar for two socialmedia scenarios. The qualitative think aloud methodol-ogy used in this work is also motivated. This is followed byan analysis of the results outlining the full spectrum of self-presentation strategies participants used during avatarcreation. This paper ends with a general discussion thatidentifies three motivating factors guiding self-presentationin social media: (1) accurate self-presentation (2) playful

self-presentation and (3) embodied messages. The theore-tical implications of these results are extended to the fieldof online communication, while several practical implica-tions are considered for the design of avatar customisationtools.

2. Background

2.1. Identity and self-presentation

Goffman (1959) considered humans to be actors,constantly following a social script. By using front,composed of setting, appearance and manner, an actoridealises his/her self-presentation as s/he conceals oremphasises aspects that correspond to the values governinga particular situation. More specifically, Goffman definedfront as the, expressive equipment of a standard kindintentionally or unwittingly employed by the individualduring his performance (Goffman, 1959, p. 32). Thesetting is the stage which the actor chooses as a backdropfor his/her performance. A first date is arranged to takeplace at a restaurant; a work meeting convenes at aconference room. The actors appearance is comprised bystable characteristics such as age, gender and race. Otheraspects of appearance are malleable such as ones displayedsocial standing or recreational activities at a given momentin time. Manner is the expressive signals one intentionallyor unintentionally gives off. For instance, a speakerpresenting at a seminar, as expected, will display a calmand confident demeanour. Conversely, despite efforts toremain calm, the speaker may give off signals of anxiety.Generally, appearance and manner are expected to align.For example, during a project pitch, the creative director isformally dressed while his/her manner should transmitconfidence with the teams proposal even though the clientmay momentarily question his/her judgment.Unlike Goffmans microscopic analysis of self, Higgins

(1987) approached the same concept from a macroscopicview. According to Higgins, there are three self-aspects,perceived either from the actors point of view or from animagined audiences perspective. The ideal self is composedof qualities which the actor or others wish for him/her topossess, while the ought self comprises qualities which theactor or others expect him/her to possess due to obligation.The actual self is composed of stable qualities the actor orothers believe s/he possesses.Goffmans and Higgins frameworks have been used to

answer questions regarding online identity and self-presentation. Aligned to Goffmans view, Ellison et al.(2006) showed that online daters have a tendency to idealisetheir self-presentation by controlling the cues they give off.In the wider field of computer-mediated communication,Bargh et al. (2002) found that online users, liberated byanonymity, generally express more actual self-aspects thanwhen they communicate face-to-face. Identity constructionin this previous work was limited to either text-based orpictorial representations. Avatars present two distinctive

ARTICLE IN PRESS

2The qualitative differences between social media and virtual environ-ments are also understood within Schroeders (2002) sociology-inspiredframework for virtual environments: an online interaction is uniquelyframed within a context and it is characterised by the social roles users arecompelled to take. In addition, the technological features of theenvironment can encourage or discourage certain behaviours by influen-cing users object of focus.

A. Vasalou et al. / Int. J. Human-Computer Studies 66 (2008) 801811802

properties, the opportunity for controlled expressivity andthe presence of (and possibility for) ambiguity. Theremainder of this section elaborates on how these propertiesmay further influence users self-presentation.

2.2. Controlled expressivity

In face-to-face communication, the actor executes theperformance in part by using his/her body as a tool,intentionally coordinating appearance and emotion expres-sion (Goffman, 1959). Though expression may constitutean intentional signal to others, it is also experienced as asymptom in the body that generates it (Hess, 2001).Expressing happiness at a wedding reflects the appropriatesocial script, while it simultaneously provokes arousal inones body. At the same time, expression is not under theactors complete control as it often follows a subjectiveemotion experience. After a social blunder, the experienceof embarrassment is spontaneously (as opposed tostrategically) expressed by blushing or by gaze avoidance(Keltner and Buswell, 1997; Tracy and Robins, 2004).Avatars are a pictorial representation of the body.

Symptoms and symbols are not interlocked into onebiological system. Because of this, an avatars signalsmight be created by its owner to align to his/her intendedexpression without spontaneously activating the corre-sponding emotion. When chatting online, one may laugh atanothers spelling error privately, but may easily choosenot to display the emotion of amusement via his/heravatar. This possibility also opens up more opportunitiesfor misrepresenting oneself, as an untalented actor nolonger needs to rely on his/her poor expressive talent, buton the in-built expressivity of the avatar. An avatar mayreflect a happy demeanour despite its owners disgruntledpersonality. Moreover, online, there is an indefinite timewindow between the act of identity construction and theactual display of front which offers the user time to reflecton the precise qualities he/she wants to emit (Ellison et al.,2006). These qualities put together can contribute to self-presentation that is exceptionally strategic and tailored toconvey precise messages.

2.3. Ambiguity

Avatar customisation environments are populated withoptions of fantasy, e.g. science-fiction costumes, as well asconventional and familiar real-life objects, e.g. jeans and at-shirt. Often, these contrasting items are equally accessibleand prominent in the interface. By contrast, in the physicalworld, an actor selects a setting for the performance thatalready exists in the real world, while clothes fit for theoccasion are worn. Rarely is the actor confronted withfictitious items for constructing his/her front such as amagic wand or an alien mask. Therefore, unlike thephysical world, avatars introduce an element of ambiguity.Ambiguity and the space it opens up for interpretation

has become an important design goal in HCI, especially for

social systems whose purpose is to provoke interpersonaldialogue (Gaver et al., 2003; Boehner and Hancock, 2006).Regarding avatars, the continuum of real and fantasyinvites users to reflect on their social roles, reinventing theirsocial preconceptions. For example, an avatar may be usedto challenge its owners traditional family role. This self-reflection can transcend into real life, shaping his/her futureencounters with family members (Turkle, 1995; Taylor,2002). Furthermore, the user can become a designer usingthe avatar as a tool to provoke the viewer; inconsistenciesand metaphors built into the avatar may create a narrativethat challenges the viewers interpretation.

3. Methods

The present research strived to understand the entirebreadth and diversity of users self-presentation strategieswhen creating an avatar for social media. Thus, it wasimportant to have direct access to users thoughts duringthe process of customisation. In choosing a methodology,we drew from the field of usability which has recentlyfocused on creating novel methods for user experiencedesign and evaluation (e.g. self-assessment manikin:Bradley and Lang, 1994; experience clips: Isomursu et al.,2007; interviews: Sundstrom et al., 2007; fuzzy logicemotion modelling: Mandryk and Atkins, 2007).The methods we came across did not allow us to collect

qualitative data concurrently to users interaction with theavatar tool; hence, we borrowed a method traditionallyused in usability studies, the concurrent think aloudprotocol. The think aloud protocol was suitable for ourpurposes as it requires users to verbalise their actions andthoughts while using an interface.

3.1. Participants

The study was advertised in bulletin boards throughoutthe University of Geneva. In total, 20 participants tookpart in the think aloud study, nine of whom were male.Participants were aged between 20 and 40 years old, allwere experienced computer participants and none hadprevious experience with avatars. As the sessions wereconducted in French, only fluent French speakers wererecruited.

3.2. Procedure

An adapted version of the think aloud protocol, tailoredfor the needs of usability studies, was used to elicitparticipants verbalisations (see Boren and Ramey, 2000).When participants first arrived they were given a trainingtask instructing them to think aloud by speaking out theironscreen actions. The facilitator made sure to positionhimself as the listener who will partake little in theprotocol, while the participant was given the role of theexpert speaker. During the session, the facilitator reaf-firmed his role as the listener by occasionally providing

ARTICLE IN PRESSA. Vasalou et al. / Int. J. Human-Computer Studies 66 (2008) 801811 803

acknowledgment tokens (e.g. hmm). If the participant wassilent for too long, the facilitator prompted him/her. Twoobservers located in a room adjacent to the lab viewed andrecorded the session. The recordings were later used for thequalitative analysis.The avatar tool used in this study was Yahoo! Avatars

(http://avatars.yahoo.fr). Yahoo! Avatars gives users con-trol over their avatars body with variable facial features,eye colour, hairstyle and colour, skin tone, clothes andaccessories. In addition to a neutral emotional expression,an avatars face can be assigned to four differentexpressions. Furthermore, avatars can be placed againstdifferent background locations. Other artefacts are alsooffered for the purpose of personalisation, such as pets,flags, furniture items, among others. The options offered inYahoo! Avatars cover a wide spectrum of everyday items,such as pets, hobbies and city locations, as well as fantasyoptions, such as masks and costumes. Therefore, choosingYahoo! Avatars ensured that participants choices were notconstrained by the possibly limited options offered by theapplication. Fig. 1 displays a screenshot of the Yahoo!Avatars customisation tool.Participants were presented with two scenarios for

creating an avatar which depicted events they had mostlikely encountered in the past i.e. a romantic date and abirthday postcard to a family member. Additionally, thescenarios were built on the grounds of previous research insocial media. For example, bloggers often write posts withthe purpose to record their life happenings (Herring et al.,2004; Nardi et al., 2004). Even though blogs are frequentlytargeted to an audience, sometimes writing is a solitarytask, a form of catharsis for working out ones emotions.Problems that occupy the user in daily life are expressedonline in order to resolve and to relieve the source ofemotion (Huffaker and Calvert, 2005). To capture these

dynamics, the first scenario prompted participants toexpress a particular life event via their avatar. Further-more, social media often links participants to familiaroffline contacts. For instance, communicating and recon-necting with friends is a motivating factor for using thesocial networking site Facebook (Joinson, 2008). Similarly,many bloggers pages serve as a channel for keeping intouch with their loved ones (Nardi et al., 2004). This socialcapacity of social media was captured by requestingparticipants to construct a message for a family member.Table 1 displays the two scenarios.Participants took 8.8min on average to complete the first

task and spent an average of 7.6min on the second task.The minimum time spent on a task was 3.4min, while themaximum was 17.45min. A paired t-test on task comple-tion time was non-significant. Therefore, participants spentabout equal time on both tasks.

3.3. Qualitative protocol analysis

During the think aloud session, participants face andvoice was recorded, as was their onscreen activity. The twoobservers of the sessions revisited these recordings toidentify moments relevant to participants chosen self-presentation strategies. The context of the interaction,participants verbal reports and participants non-verbalexpressions, e.g. smiling, were all consulted when extract-ing relevant incidents. Usability problems with the inter-face, e.g. this button does not work, were discarded. Thisprocess resulted in a total of 202 incidents which were thentranscribed and translated in English. Next, the incidentswere grouped by similarity forming four high-levelcategories of self-presentation strategies. These are: stableself-attributes, idealisation, messages to others and fun. Inthe next section, we analyse the results by categorising theincidents according to this classification.

4. Results

4.1. Stable self-attributes

Participants expressed a number of stable self-aspectsthat were not influenced by the two scenarios. These

ARTICLE IN PRESS

Fig. 1. Yahoo! Avatars customisation tool.

Table 1Social media scenarios

Scenario 1: You are going out on a first and much anticipated romanticdate. There are many questions you have to plan for. Where should wemeet? What should I wear? How friendly should I look? What should Italk about? You decide to display your date online via your Yahoo! avatarScenario 2: Your moms birthday is coming up and you have decided tosurprise her with a holiday at a Greek island. You have orchestrated allthe details: the perfect season for good weather, delicious local foodguaranteed and a quiet hotel room by the seashore. Instead of telling herin person, you want to surprise her by presenting your vacation offer in amore fun and unexpected packaging. You will send a postcard withYahoo! Avatars

A. Vasalou et al. / Int. J. Human-Computer Studies 66 (2008) 801811804

were: (1) personalising a face, (2) fashion statements, (3)values and (4) real life events.

4.1.1. Personalising a faceYahoo! Avatars provide eight face choices, each of

which displays different characteristics. Generally, theaccurate reflection of oneself via a face was an importantchoice, and, thus, one worth the time investing in. Fiveparticipants chose a face at the onset, only to revisit theeight faces later, in search of a closer match to their actualvisage. Unlike other avatar items that were malleable andcontext-driven, once chosen, the face remained constantacross the two scenarios. When approaching a face,participants either referenced themselves or consideredthe avatar as an external agent with a character of its own.We consider each strategy separately.People define themselves in different ways, which in turn

can be projected onto their avatar. For example, onesidentity can be linked to a particular facial characteristic.A testament to our ability to recognise distinctive featuresand use them as identity cues are the range of adjectives weuse to classify people we know e.g. he has an oval face,she has alabaster skin. To that effect, one of ourparticipants, while selecting a face, said:

y well, I do not find eyes which resemble miney Theface does not resemble too much my own (sighs) but itwill doy

This participant defined himself by the uniqueness of hisalmond-shaped eyes. He attributed the lack of resemblancebetween himself and his avatar to the mismatch in eyeshape completing his choice with a feeling of disappoint-ment. Participants following this style looked for definingpersonal features, such as a particular facial shape or abeard they wore. The feeling resulting from ones inabilityto accurately represent his/herself was disappointment.Alternatively, identity may instead be linked to ones

understanding of his/her own characteristics. Here,mood and disposition as expressed by the avatar facebecame more important than the similarity with partici-pants physical facial features. Although the eight avatarfaces available were given numbered labels by Yahoo!designers, some participants assigned them emotional

meaning (see Fig. 2). In the following extract, one of ourparticipants, a Caucasian female, chose an Asian face onthe basis of its mood. The avatar was perceived as happyand thus reflected her state of mind, even though it did notaccurately represent her own physical appearance.

yfaces are different, with different emotions; I willmaybe choose a face with a little smiley

In the examples given so far, participants sought toexpress their own perceived qualities and characteristicsthrough their avatar. Hence, the avatars face should reflectcertain aspects of its owner. However, people often alsoassociate personality traits to a face in the absence of anyother cues (Taylor, 2002). For instance, one participantjokingly described the avatars faces thus:

yone avatar looks rather well-behaved (laughs); thisone looks like a playboy (laughs)y

In this context, avatar faces were treated as independentagents with a distinct personality and correspondingbehaviours. In the above example, the participants finalchoice was to choose an avatar face that resembled hisown. Thus, one strategy does not preclude another and afew participants vacillated from one strategy to the other.In addition to the faces available, participants could

personalise other facial attributes such as the eye colour,hairstyle and colour, and skin tone. Unlike faces, theseadditional attributes were only partially enduring. Partici-pants reflected their natural eye and hair colour, rarelyrevising these two choices. A similar observation was madefor the skin tone as well as the hairstyle. For instance, asone participant noted I will first select the colour of theskin. I have light skin so I will keep it this way. However,during our analysis, we also discovered that skin toneand hairstyle were often malleable features driven bystrategic motivations. This particular observation isanalysed in-depth in Sections 4.2 and 4.3.

4.1.2. Fashion statements4.1.2.1. Clothes express personality. People have particu-lar ideas when it comes to considering the kind ofperson they are. One may think of oneself as a matureperson, while another may want to project youthfulness.

ARTICLE IN PRESS

Fig. 2. Male and female avatar faces.

A. Vasalou et al. / Int. J. Human-Computer Studies 66 (2008) 801811 805

A participant wearing eye wear in real life searched forglasses. He frowned when viewing the options availablenoting the disparity to his self-image: No, the glasses looktoo young. Another participant was unsuccessful infinding glasses to match his real life glass frame. As heexplained, Here it could be good to vary more, to havevarious types of glass frames. In both instances,participants opted out of the eye wear option. In a similarvein, while one participant embraced the designersdecision to include male jewellery allowing him to reflecta real life piercing, another participant criticised malejewellery as being too feminine.

4.1.2.2. Combinations should reflect good taste. Clothes,shoes and accessories were chosen to match in both styleand colour. As in real life, when choosing combinations,participants were careful to reflect good taste. Exemplify-ing this, one of our participants exclaimed in disgust upontrying on a blue Hawaiian shirt and black Hawaiiantrousers Yuck! (laughs) That does not match well.Another participant who was looking for glasses to matchwith his outfit said:

yI would like to find something which will correspondwell to what I am wearingy Glasses, black frames(searching) therefore I should choose this.

This participant felt compelled to take the mostcomplimentary option. His choice was motivated by styleand, thus, glasses that were incongruent with his outfit werenot favourable.

4.1.2.3. Got-to-have items. I am a shoe person;Earrings are my trademark. Statements of passion aboutparticular clothing items are expressed all the time. Fashionfavourites were sought after when creating online avatars.One participant, a self-reported shoe lover, exclaimed:

Are there shoes? Is my avatar wearing shoes? Can onechange the shoes?y Because shoes are important to me.But, there seems to be no such option (sigh).

In Yahoo! Avatars, shoes come automatically withcorresponding clothes and are one of the few options thatare not customisable. Yet, this participants disappoint-ment was echoed throughout the study by eight others whorepeatedly searched for ways to customise their shoes.Crucially, a few participants cancelled their clothingoptions when the corresponding shoes were not to theirliking. In a similar vein, some female participants lookedfor makeup, a male participant searched for a piercingsimilar to his own.

4.1.3. Values4.1.3.1. Group affiliations. San Francisco, Venice andMoscow are a few of the worldwide destinations ondisplay. In the following example, a participant firstsearched for Swiss clothes. He said doubt that there areSwiss clothes, which is quite a pity! Shortly after, he

turned his attention to the backgrounds available andsearched again with the same purpose: On the other hand,I have the impression that those are backgrounds so ifthere is one of Switzerland it could be nice! Thisparticipant wanted to express his cultural belonging viahis avatar. When realising there were no Swiss optionsavailable, he abandoned his search with disappointment.Another, Mexican participant drew on his culturalupbringing when conveying the surprise holiday to hismom; instead of a Greek island, he chose to displayAcapulco.

4.1.3.2. Moral principles. Yahoo! offers a branded itemssection comprised by clothing and accessories frompopular brands, which can be used to dress ones avatar.This new generation of advertising is inconspicuouslyembedded in the avatar environment (Hemp, 2006), but itdoes not come without criticism. While looking at thebranded items, five participants expressed their oppositiontowards advertising and freedom of choice. As oneexplained:

So, this really makes me think that Yahoo! Avatars canbecome an advertising tooly prime people to getinterested into the branded stuff. That is not somethingthat I like much.

In this instance, participants ideological viewpointsnegatively affected their self-presentation choices and theirtrust in Yahoo! As opposed to this negative outlook ononline advertising, another participant cast a different viewon brand advertising: brands do not interest me, except ifthey have Decathlon (his favourite brand). It is thuspossible that inhibitions are bypassed when advertising ispersonalised with the inclusion of favourite offline brandnames.

4.1.3.3. Emotional attachments. Sports fans ferventlyvouch for their team. Motorcycle riders take yearly grouptours to show their solidarity to one another. Hobbiesrequire both a positive emotional bias and time commit-ment. Yahoo! Avatars provides many choices for showcas-ing ones hobbies, which in turn appeared to be welcomedby participants. An ice skater chose an ice ring; a baseballfan dressed up in uniform; a motorcycle rider placed amotorcycle next to his virtual self. A hockey fan whoselected a hockey stadium background said good, theyhave hockey because I practise it; I adore hockey (laughs).Making this choice shows a little of my personality.Another object of emotional attachment was animals.Animals require time commitment and constant care, butin return they offer us friendship and companionship. Weattribute personality to animals which we learn to acceptand love. The importance of this emotional bond was alsoevident in this study when several participants chose theirreal life companions as their online companions. As oneparticipant noted I am looking for a dog that actuallylooks like my dog which is a Chihuahua. Another

ARTICLE IN PRESSA. Vasalou et al. / Int. J. Human-Computer Studies 66 (2008) 801811806

participant selected a cat when customising her avatar forthe birthday scenario as she always takes her cat with herwhen on holiday.

4.1.4. Real life eventsCustomisation options present in the site were used to

express real life occurrences. For example, a few partici-pants chose darker skin tone to reflect the summer season.Another participant expressed his recent visit to the barberwith the corresponding avatar option: Good, today I wentto the hairdresser, therefore, I choose this one: very shortbrown hair. Though the study took place during thesummer, when attending on a rainy day, one participantchose long pants and a hat to keep himself protected fromthe weather.

4.2. Idealisation

In Goffmans (1959) terms, participants idealised theirself-presentation by intentionally emphasising or conceal-ing self-attributes. We found three patterns of idealisation:(1) global idealisation, (2) social idealisation and (3)idealisation versus breaking the rules.

4.2.1. Global idealisationSeveral participants slightly morphed their avatar to

display an ideal and more attractive self (Higgins, 1987).For example, avatars were given a darker skin toneexpressing their owners desire for a more vibrant summerlook. One participant acknowledged this embellishment bysaying I will pretend to look more suntanned. Further-more, a male participant with a receding hair line in reallife started off with the intention to accurately reflect hishairstyle. Instead, he was confronted with hair options attwo ends of a spectrum, either a completely bald head or ahead with short hair. He negotiated these two possibilities:

Lets try something with no hair, especially in theforehead. That looks close (while viewing the shorthair). Not as much, maybe in a few years from now(laughs) (while viewing the bald head).

Even though this participants initial intention was totruthfully depict his appearance, he found himself caughtbetween two extremes: his past self and possibly future self.His final choice depicted an idealised and more youthfulappearance, chiefly imposed by the limited hair choicesoffered within the application.

4.2.2. Social idealisation4.2.2.1. Romantic partner. In the dating scenario, parti-cipants predominantly enacted the role they usually playwhen impressing a romantic partner. Setting, appearanceor manner was altered accordingly. In some extreme cases,all three facets of front were used in concert for theperformance. Specifically, many participants chose back-grounds of conventional locations for a romantic meetinge.g. a dinner evening in a terrace, a date at a disco. Others

searched for a setting that depicted particular traits. Togive an example, one participant selected a tranquilhiking location, just right for getting to know someonenew. Participants searched for elegant clothes: severalfemale participants selected evening dresses, while maleparticipants selected suits or formal pants. Other props forbuilding up an attractive self-presentation were used suchas jewellery, handbags or makeup. Finally, a few partici-pants selected a happy expression for their avatar, suitablefor a romantic evening. Unlike global idealisation whichwas guided by ones personal expectations (see Section4.2.1), in this instance an ideal self was oriented to theexpectations of a romantic partner.

4.2.2.2. Broadcaster of happy news. In the birthdayscenario, a few participants morphed their self-presentationto align to the role of the happy broadcaster. A maleparticipant chose a guitar to announce the happy news:I will choose the avatar with the guitar. I could maybesing something for my mom; happy birthday, of course!A female participant took the role of the messenger bydressing her avatar with festive clothes accompanied by ahappy emotion expression:

I am supposed to announce the gift so I need to wearsomething special (chooses a skirt). There we go, warmcolours (chooses an orange shirt). Happy, sad, special(browsing through the avatars emotions). We will gowith happy. Its perfect.

One participant had browsed through the availableemotion expressions, while completing the dating scenario.He said: special, what is the special emotion? O no! This islousy, it is a losers looky Sad emotion, no not sadyHappy emotion, this is not happy, it is idiotically happy.However, when personalising the avatar with his mom inmind, he noted since this is for my mom, I will put up asmile even though I will look stupid. Though the avatarsemotions were not aligned to his expressive preferences,when announcing happy news he abandoned his inhibi-tions and selected a happy expression.

4.2.2.3. Child. While completing the birthday scenario,some participants drew on the role of the child who isobliged to conform to an ought self (Higgins, 1987).Specific features in their appearance were concealed toalign to their parents expectations. For instance, a femaleparticipant browsed through several swimming suits beforechoosing the most conservative less naked choice. In thefollowing extract, a participant who had enthusiasticallychosen a piercing similar to his own for the date scenarioremoved the piercing while explaining: I will take off thepiercing; otherwise my mom will tell me that I am scum(laughs).

4.2.3. Idealisation versus breaking the rulesEven though most participants drew on existing social

roles during the process of customisation, avatars also


encouraged breaking the rules as contrasting items couldbe easily combined into one composition with as little asone click. A female participant chose sophisticated Parisas her destination, while wearing a cowboy outfit she hadearlier selected for her date. She hesitated for a moment butthen embraced her inappropriate choice: here we go(selected Paris); am I wearing clothing suited for down-town? Not really, actually, (pause) but well, I like Paris, letus go! (laughs). A male participant who had chosen achic suit purposefully for the date, browsed throughbackgrounds in search of a romantic location. His finalchoice was a farm. He noted the incongruence between hisidealised appearance and the background: the farmbackground is not really aligned to the suit (and thus tothe date), but it matches the style of the hat. Despitesetting out to personify the romantic partner, thisparticipant was able to quickly and easily dismiss thesocial conventions he had initially chosen to follow.

4.3. Messages to others

4.3.1. Literal versus symbolic messagesIn the birthday scenario, similar to what was observed in

the social idealisation strategy (see Section 4.2.2) partici-pants orchestrated setting, appearance and manner duringthe avatar construction. However, instead of adapting aparticular social role oriented towards an imaginedaudience, all but two participants enacted their modelbeach holiday. Therefore, the avatars appearance wasmanipulated to send a literal message to a particularrecipient. Beach backgrounds were taken that wereindicative of a Greek island. Avatars were dressed inbathing suits or in light summer garments. Props wereused to further highlight the holiday: some participantschose sun glasses; others chose summer hats; and twofemale participants chose summer handbags. Severalparticipants coloured their skin tone darker to furtherillustrate the beach holiday, whereas one participantlightened her skin tone to express the fact that she hadnot yet left for the holiday. Finally, two female participantschose more casual hairstyles suitable for the beach. Asone explained: I would like a haircut for the holidays.The small ponytail is no longer fit. Oh, dreadlocks aregood.Even though avatars were predominantly used to

communicate literal messages, the abundant visual optionswithin the site also led to the construction of somesymbolic messages. One participant used imagery to attachinformation of significance to the message recipient. Hesaid my mom likes animals a lot and cats in particular.In orienting his avatar towards another, he searched for acat companion. Another participant used the backgroundimage to construct a metaphor: a black background filledwith light bulbs communicated her brilliant holiday idea.Yet, a third participant chose a birthday party backgroundwith a cake and gifts to signify the festive birthday spirit.

4.3.2. Provoking a reactionSeveral participants explicitly morphed their avatar to

provoke an emotional reaction from the message recipient.In the following two examples, the sender envisioned thereceivers response, experiencing positive empathy for theother and assigning even more emotional value to theircustomisation choice. A participant who had earlierselected a self-representative hairstyle now said I wouldlike to change the haircut for the beach. I want to findsomething funny for my momy Afro hairy This one isfunny (smiles). In a similar vein, a second participant gavehis avatar a suit even though his taste in fashion wascasual. He explained his moms reaction if she were to viewhis avatar: She will be proud of the little boyyMy avataris smartly dressed; that will give her a shock (laughs).

4.4. Fun

Finally, some participants gravitated towards optionsthat were simply fun and entertaining to them personally.One participant chose an amusing hat because it remindedhim of Indiana Jones. Another participant chose a dogsimilar to her own which was dressed in a raincoat.Laughingly, she commented twice on her dogs comicalappearance. Many others assigned their avatar an ani-mated emotion because of its entertainment value. In thefollowing example, a participant browsed through theemotions sad, angry and special, in the endselecting the special emotion:

Excellent! Yes! (selects the sad emotion) Ohhhh, he issad, he is very sad (laughs). (selects the angry emotion)This one is angry, not bad. (selects the special emotion)And the special one, yes the special one, it is cool thespecial one (laughs). Let us zoom into see, do again thatlittle thing you did earlier, yeah, he is cool my guy(laugh)y

5. Discussion

The discussion that follows organises the self-presenta-tion strategies delineated in Section 4 into three motivatingfactors driving self-presentation: (1) accurate self-presentation,(2) playful self-presentation and (3) embodied messages.We acknowledge that the three motivating factors do notnecessarily operate independently; while an avatar usermay reflect an accurate self, elements of playful self-presentation may be additionally chosen.

5.1. Accurate self-presentation: mirroring ones offline selfonto an avatar

Participants reflected their actual self (Higgins, 1987) byselecting options aligned to their perceived stable self-attributes. Across both scenarios, avatars physical char-acteristics were chosen carefully to reflect their ownersappearance. Fashion preferences that guided participants


appearance in the physical world, also determined theironline choices. Moreover, participants drew on a largerframework of stable values when personalising their avatarincluding cultural affiliations and moral beliefs.In addition to expressing an actual self, participants

idealised their avatar by concealing or emphasisingparticular self-attributes. First, social roles determinedhow participants chose to present themselves (Goffman,1959). When expressing an impending romantic date, manyusers constructed avatars that looked more attractive,while backgrounds depicted popular romantic destinations.Therefore, an ideal self was prevalent, composed ofqualities desired by a romantic partner. When constructinga message for their mom, a few participants projected anought self (Higgins, 1987) by intentionally creating a moreconservative appearance. Some participants also adoptedthe role of the broadcaster, whose aim was to deliver goodnews with the display of a happy and festive fac-ade. Whenparticipants constructed the birthday surprise, even thoughthe avatar itself constituted the message (see Section 5.3),they drew together all the elements that described aholidaymaker. Second, some participants idealised theiravatar with slight improvements they personally desired(Higgins, 1987). Nonetheless, this type of intentionalembellishment was not a predominant trend in this study.As originally proposed in Section 2.2, we found that

participants composed their avatar to give off precisemessages that intentionally avoided ambiguity. A femaleparticipant chose an evening dress, to later notice a non-representative music sheet in the hands of her avatar. Sheopted for another dress even though it was less congruentto her fashion style. Similarly, a male participant searchingfor a suit aimed to display elegance without diluting hisself-image with other attributes: I will choose this suit as itis more elegant. The other one looks too much like abusinessman. However, even though participants care-fully selected options that were aligned to an imaginedsocial role, they tended to be reflective of their stable self-attributes and preferences (see Section 4.1). As a femaleparticipant stated: the first date is important for me. I willtry to find a way of dressing nicely, but that also reflectsme. I am looking for something that looks like a student.This one may be too young. I would like somethingpleasanty This one reflects me so I save. Precise controlover ones idealised self-presentation and the need for self-representative options were competing factors that directedparticipants choices either closer to an actual self or tostrategies of idealisation. In an illustrative example, oneparticipant had earlier chosen glasses with a frame similarto his own. When completing the birthday scenario, helooked for sunglasses. He searched in vein for a pair heliked, finally deciding to keep his original eye wear.Overall, the above conclusion is in line with observa-

tional data suggesting that users of social media presentthemselves accurately (Herring et al., 2004; Vasalou et al.,2007). In this study, accurate self-presentation was builtinto avatars either in the form of stable self-attributes or in

the form of idealisation. To achieve this, participantswelcomed the spectrum of choices made available byYahoo! designers for setting, appearance and manner.Hobbies, pets and other self-relevant props were allimportant elements for identity construction (Goffman,1959). Thus, alongside the control offered over ones onlineembodiment, designers of avatar customisation toolsshould put forward other artefacts that are equallyimportant to identity expression.Furthermore, when constructing ones self-presentation,

lack of flexibility over ones options ultimately determinedthe avatars final appearance and, thus, the degree ofidealisation the avatar owner expressed. This latter pointhas some implications for interpersonal communication.Yee and Bailenson (2007) found that participants repre-sented by attractive avatars were more willing to approachmembers of the opposite gender, as compared to partici-pants represented by less attractive avatars. In a negotia-tion task, participants with taller avatars were moreconfident than participants represented by shorter avatars.By preventing or enabling the display of particular avatarattributes, avatar designers exert influence over usersidealisation, in turn possibly shaping users online beha-viour (also see Taylor, 2003).

5.2. Playful self-presentation: reinventing oneself and sociallife

In Section 2.3, it was suggested that some avatarcustomisation tools present equal availability and promi-nence of fictitious objects and also objects found in thephysical world. This introduces a level of ambiguity that inturn challenges ones own self-conception, but also onessocial orientation towards another. Aligned to this predic-tion, we found that the ambiguity fostered in Yahoo!Avatars led to five types of strategies. Participantsabandoned their preconception of social correctness andexperimented by mixing inappropriate choices together.Therefore, social rules that steer ones self-presentationoffline were bent and circumvented online. Participantsalso used the avatar to build narratives about events takingplace in their physical world. As opposed to reporting onespresent state via a textual channel, as offered in most socialmedia platforms, e.g. blogs, social networking applications,participants used the avatars body as a means forexpressing their ongoing life events. Furthermore, someconstructed symbolic and metaphorical messages thatinvited the viewers interpretation. Additionally, partici-pants manipulated their appearance in order to provoke apositive emotional reaction from the avatar viewer. Finally,selections were made that were non-representative ofparticipants offline self, but were chosen because of theirentertainment value.As discussed in Section 5.1, many participants aimed to

construct a precise self-presentation, as a result criticisingthe avatar application over its lack of flexibility. However,the presence and diversity of predefined (and thus less


customisable) objects also created a new space for creativeself-expression and communication. Participants whoexploited the ambiguity afforded by the avatar customisa-tion tool advanced their identity beyond the mere reflectionof an offline self. The present result is also in line withobservations from virtual environments; to give anillustrative example, in a study reported by Taylor (2002),a female avatar user distanced herself from users she wasnot familiar with, symbolically, by wearing a witch hat.Hence, even though computational resources may ulti-mately allow users to shape their online representation withprecision, a strong argument can be made in favour ofambiguity and the user engagement it fosters. We thusargue that designers of avatar environments shouldcontinue maintaining a predefined palette of diversecustomisation options.

5.3. Embodied messages: avatars as proxies

When announcing a surprise holiday gift on the occasionof their moms birthday, participants enacted their idealholiday break via their avatar. Beach backgrounds werepreferred, while bathing suits and other holiday items wereselected. By using front, participants expressed theirholiday self in a similar manner to those who engaged insocial idealisation. However, in this instance, ones digitalrepresentation embodied a literal message towards another.Researchers and practitioners have previously focused

on creating avatars with in-built emotion expressivity inorder to enable the exchange of richer emotional content.For example, in the V-Chat virtual environment, userswere given the ability to compliment their written messageswith static emotional gestures (Cheng et al., 2002). Fabriand Moore, 2005 built a chat messenger which allowedusers to change their avatars facial emotion expression.This kind of approach naturally follows from our intrinsiccommunicative mechanisms. However, the present resultindicates that avatar users can invent new and unconven-tional communicative gestures such as using the avatarsappearance to embody the message.Future work could investigate users communicative

intentions in order to reveal what kind of customisationoptions are needed to support embodied messages.Additionally, further research might examine whetherembodied messages influence communication at the inter-personal level. The expression and interpretation ofemotions is frequently socially and culturally constructed.A reciprocal emotional exchange can occur in which thesenders emotion expression elicits an emotion in thereceiver (Parkinson, 2008). Aligned to this view, Fabri andMoore (2005) found that users chatting online mirroredeach others emotions as expressed via their avatar. In thisstudy, we discovered that senders are happy to give up theirneed to depict an actual self in favour of reaching out andengaging with another user. Converting ones digitalrepresentation into an expressive signal may be perceivedby the message recipient as an altruistic gesture. As a

consequence, it is likely that embodied messages facilitateintense emotional exchanges between message sender andreceiver, even more so than the exchange of fleetingemotion expressions.

5.4. Limitations

To our knowledge, this study is the first attempt toanalyse users self-presentation strategies when creatingavatars for social media. The concurrent format of thethink aloud method allowed us access to participantsthoughts in order to gain this fine-grained knowledge.However, this methodological choice also placed partici-pants outside their natural environment and inside the lab;the social media scenarios motivating the avatar creationcannot capture the entire spectrum of participants inten-tion of use. It is important that future research corrobo-rates and extends our results by ethnographic means.Moreover, even though the facilitator was trained not tointerfere during participants avatar creation, we recognisethat the observation methods used in this study may haveprevented some participants from verbalising particular,perhaps more embarrassing, self-presentation strategies.Finally, Yahoo! frequently updates the Yahoo! Avatarscustomisation tool with new selections. As the site grows,users may continue inventing new strategies for self-presentation.

6. Conclusion

The present research set out to investigate how userschoose to present themselves via an avatar for social media.Using the concurrent think aloud protocol, three distinctmotivating factors in avatar construction were found.First, participants depicted accurate self-presentations thatwere aligned either to an actual self, as expressed by onesperceived stable self-attributes, or to an ideal or ought self,as expressed by strategies of idealisation. Stable self-attributes and idealisations are competing strategies, oftenresolved by the availability of options within the customi-sation tool. Avatar designers may want to acknowledgeusers need for constructing precise self-representations byoffering additional control and flexibility over customisa-tion choices. Further, while creating an accurate self-presentation, participants drew together a collection ofobjects that depicted setting, appearance and manner. Onthese grounds, we suggest widening the scope of avatarcustomisation to include options that are not necessarilyattached to the avatars body but are yet central to usersidentity (e.g. pets). Second, ambiguity allowed participantsto create playful self-presentations. Some participants usedthe avatar to challenge the social conventions that usuallydirect self-presentation when offline. Others built narra-tives into the avatars body that reflected life events or thatdirectly engaged the viewer. Thus, even though avatarapplications may have the future capacity to allow thecreation of precise and photorealistic representations,


designers should continue to support ambiguity byadditionally providing pre-constructed customisation op-tions. Third, avatars were used to embody a message asparticipants oriented their entire self-presentation toconvey a literal message to another user. Therefore, theavatars body can be converted by the user into anexpressive signal. Future research can advance the designof avatar customisation tools by gaining a deeper under-standing on the occasions that give rise to embodiedmessages. However, this result also opens up new questionsfor online communication. Orienting ones entire bodytowards another may be perceived by the viewer as analtruistic and generous gesture, in turn facilitating theexchange of highly emotive messages.In conclusion, the above findings advance the topic of

identity in social media by revealing the strategies thatusers bring together when creating an avatar. Moreover,they suggest several implications for the design of avatarcustomisation tools and more generally for online com-munication.

Acknowledgements

We thank Jarmo Laaksolahti (SICS) and Kristina Hook(KTH/SICS) for their constant support during thisresearch. Etienne Roesch (University of Geneva), SophiaVasalou (University of Cambridge) and Marc Fabri (LeedsMetropolitan) are acknowledged for their comments on anearlier draft of this paper. We are thankful to threeanonymous reviewers whose comments improved thepaper. Steve Binggeli, Delphine Messot and EtienneRoesch assisted with the data collection and the transcrip-tions. This research was funded by the HUMAINE ISTFramework VI Network of Excellence.

References

Axelsson, A.S., 2002. Digital dividestatus differences in virtualenvironments. In: Schroeder, R. (Ed.), The Social Life of Avatars:Presence and Interaction in Shared Virtual Environments. Springer,London, pp. 188204.

Bargh, J.A., McKenna, K.Y.A., Fitzsimons, G.M., 2002. Can you see thereal me? Activation and expression of the true self on the Internet.Journal of Social Issues 58, 3348.

Bers, M., Gonzalez-Heydrich, J., DeMaso, D., 2001. Identity constructionenvironments: supporting a virtual therapeutic community of pediatricpatients undergoing dialysis. In: Proceedings of the CHI. ACM Press,New York, pp. 380387.

Boehner, K., Hancock, J.T., 2006. Advancing ambiguity. In: Proceedingsof the CHI. ACM Press, New York, pp. 103106.

Boren, M.T., Ramey, J., 2000. Thinking aloud: reconciling theory andpractice. IEEE Transactions on Professional Communication 43,261278.

Bradley, M.M., Lang, P.J., 1994. Measuring emotionthe self-assessmentmannequin and the semantic differential. Journal of Behavior Therapyand Experimental Psychiatry 25, 4959.

Bruckman, A., 1993. Gender Swapping on the Internet. In: Proceedings ofthe INET.

Cheng, L., Farnham, S., Stone, L., 2002. Lessons learned: building anddeploying shared virtual environments. In: Schroeder, R. (Ed.), The

Social Life of Avatars: Presence and Interaction in Shared VirtualEnvironments. Springer, London, pp. 90111.

Ellison, N., Heino, R., Gibbs, J., 2006. Managing impressions online: self-presentation processes in the online dating environment. Journal ofComputer-Mediated Communication 11.

Fabri, M., Moore, D.J., 2005. Is empathy the key? Effective communica-tion via instant messaging. In: Proceedings of the 11th EATAInternational Conference on Networking Entities.

Gaver, W.W., Beaver, J., Benford, S., 2003. Ambiguity as a resourcefor design. In: Proceedings of the CHI. ACM Press, New York,pp. 233240.

Goffman, E., 1959. The Presentation of Self in Everyday Life. Doubleday,Garden City, NY.

Hemp, P., 2006. Avatar-based marketing. Harvard Business Review.Herring, S.C., Scheidt, L.A., Bonus, S., Wright, E., 2004. Bridging the

gap: a genre analysis of weblogs. In: Proceedings of the 37th HawaiiInternational Conference on System Sciences (HICSS-37).

Hess, U., 2001. The communication of emotion. In: Kaszniak, A. (Ed.),Emotions, Qualia, and Consciousness. World Scientific Publishing,Singapore, pp. 397409.

Higgins, E.T., 1987. Self-discrepancy theory. Psychological Review 94,11201134.

Huffaker, D.A., Calvert, S.L., 2005. Gender, identity, and language use inteenage blogs. Journal of Computer-Mediated Communication 10.

Isomursu, M., Tahti, M., Vainamo, S., Kuutti, K., 2007. Experimentalevaluation of five methods for collecting emotions in field settings withmobile applications. International Journal of HumanComputerStudies 65, 404418.

Joinson, A., 2008. Looking at, looking up or keeping up with people?Motives and uses of facebook. In: Proceedings of the CHI. ACMPress, New York, pp. 10271036.

Keltner, D., Buswell, B.N., 1997. Embarrassment: its distinct form andappeasement functions. Psychological Bulletin 122, 250270.

Mandryk, R.L., Atkins, M.S., 2007. A fuzzy physiological approach forcontinuously modeling emotion during interaction with play technol-ogies. International Journal of HumanComputer Studies 65, 329347.

Nardi, B.A., Schiano, D.J., Gumprecht, M., Swartz, L., 2004. Why weblog. Communications of the ACM 47, 4146.

Nowak, K.L., Rauh, C., 2005. The influence of the avatar on onlineperceptions of anthropomorphism, androgyny, credibility, homophily,and attraction. Journal of Computer-Mediated Communication 11.

Parkinson, B., 2008. Emotions in direct and remote social interaction:getting through the spaces between us. Computers in Human Behavior24, 15101529.

Schroeder, R., 2002. Social interaction in virtual environments: key issues,common themes, and a framework for research. In: Schroeder,R. (Ed.), The Social Life of Avatars: Presence and Interaction inShared Virtual Environments. Springer, London, pp. 118.

Sundstrom, P., Stahl, A., Hook, K., 2007. In situ informants exploring anemotional mobile messaging system in their everyday practice.International Journal of HumanComputer Studies 65, 388403.

Taylor, T.L., 2002. Living digitally: embodiment in virtual worlds.In: Schroeder, R. (Ed.), The Social Life of Avatars: Presenceand Interaction in Shared Virtual Environments. Springer, London,pp. 4062.

Taylor, T.L., 2003. Intentional bodies: virtual environments and thedesigners who shape them. International Journal of EngineeringEducation 19, 2534.

Tracy, J.L., Robins, R.W., 2004. Putting the self into self-consciousemotions: a theoretical model. Psychological Inquiry 15, 103125.

Turkle, S., 1995. Life on the Screen: Identity in the Age of the Internet.Simon & Schuster, New York.

Vasalou, A., Joinson, A., Pitt, J., 2007. Constructing my online self:avatars that increase self-focused attention. In: Proceedings of theCHI. ACM Press, New York, NY, pp. 445448.

Yee, N., Bailenson, J., 2007. The proteus effect: the effect of transformedself-representation on behavior. Human Communication Research 33,271290.


Avatars in social media: Balancing accuracy, playfulness and embodied messagesIntroductionBackgroundIdentity and self-presentationControlled expressivityAmbiguity

MethodsParticipantsProcedureQualitative protocol analysis

ResultsStable self-attributesPersonalising a faceFashion statementsClothes express personalityCombinations should reflect good tasteGot-to-have items

ValuesGroup affiliationsMoral principlesEmotional attachments

Real life events

IdealisationGlobal idealisationSocial idealisationRomantic partnerBroadcaster of happy newsChild

Idealisation versus breaking the rules

Messages to othersLiteral versus symbolic messagesProvoking a reaction

Fun

DiscussionAccurate self-presentation: mirroring ones offline self onto an avatarPlayful self-presentation: reinventing oneself and social lifeEmbodied messages: avatars as proxiesLimitations

ConclusionAcknowledgementsReferences

avatars in social media

Documents