Collaborative Virtual Environment Technology for People With Autism

http://foa.sagepub.com/Developmental DisabilitiesFocus on Autism and Other

http://foa.sagepub.com/content/20/4/231The online version of this article can be found at:

DOI: 10.1177/10883576050200040501

2005 20: 231Focus Autism Other Dev DisablDavid Moore, Yufang Cheng, Paul McGrath and Norman J. Powell

Collaborative Virtual Environment Technology for People With Autism

Published by:

Hammill Institute on Disabilities

and

http://www.sagepublications.com

can be found at:Focus on Autism and Other Developmental DisabilitiesAdditional services and information for

http://foa.sagepub.com/cgi/alertsEmail Alerts:

http://foa.sagepub.com/subscriptionsSubscriptions:

http://www.sagepub.com/journalsReprints.navReprints:

http://www.sagepub.com/journalsPermissions.navPermissions:

What is This?

- Jan 1, 2005Version of Record >>

at CORNELL UNIV on November 18, 2014foa.sagepub.comDownloaded from at CORNELL UNIV on November 18, 2014foa.sagepub.comDownloaded from

http://foa.sagepub.com/

http://foa.sagepub.com/content/20/4/231

http://www.hammill-institute.org/

http://www.sagepublications.com

http://foa.sagepub.com/cgi/alerts

http://foa.sagepub.com/subscriptions

http://www.sagepub.com/journalsReprints.nav

http://www.sagepub.com/journalsPermissions.nav

http://foa.sagepub.com/content/20/4/231.full.pdf

http://online.sagepub.com/site/sphelp/vorhelp.xhtml



Collaborative virtual environments (CVEs) hold great potentialfor people with autism. An exploratory empirical study wasconducted to determine if children and youth with autism couldunderstand basic emotions as represented by a humanoidavatar. Thirty-four participants (ages 7.8–16 years) reported tohave autism interacted with a software program designed toevaluate their ability to identify and make inferences from facialexpressions. Over 90% of the participants accurately recog-nized emotions displayed by avatar representations. Thesefindings support the optimism that CVEs can be used effec-tively as an assistive technology, as an educational technology,and as a means of helping address potential theory-of-mindimpairments.

There has been an increase recently in research interestin the use of computer-based learning (CBL) ap-proaches with people with autism. This has been mo-

tivated partly by a belief that education is central to improvingthe lives of people with autism (Aarons & Gittens, 1998) andpartly by a belief in the benefits of CBL for people with autism.Many people with autism appear to have a natural affinity forcomputers and the controlled environment provided by thecomputer, and they may benefit from individualized tutoringand repetition of exercises (Hardy, Ogden, Newman, & Cooper,2002; Moore, McGrath, & Thorpe, 2000; Moore & Taylor,2000; Neale, Cobb, & Wilson, 2002).

Such arguments have resulted in an increase in researchand development work involving CBL for users with autism(see Moore, 2005, for an overview of recent work in the field).Despite this, however, the field remains largely unexplored(Beardon, Parsons, & Neale, 2001; Hardy et al., 2002). Inparticular, one promising approach to CBL, which has thus farseen very little research as an educational tool for people withautism, is the use of “virtual environments.” The research dis-cussed in this article is chiefly concerned with the notion ofusing such environments, and in particular collaborative vir-

tual environments (CVEs), as a potentially valuable technol-ogy for people with autism.

In the next section, we briefly outline CVE technology andits central feature, the notion of an on-screen avatar. This en-ables us to present a prima facie case for the potential utility ofCVE technology as an aid for people with autism. We then re-view the limited amount of current evidence available in thisarea. Next, we outline a simple computer-based system webuilt as a vehicle for our exploratory study of the ability of userswith autism to interact with avatars, and we discuss the study.

Virtual Environments

A virtual environment (VE) can be defined as a computer-generated three-dimensional simulation of a real or imaginaryenvironment (Cobb, Kerr, & Glover, 2001). Two forms of vir-tual environment can be distinguished: a single-user VE (SVE)and a multiuser, collaborative VE (CVE; Neale et al., 2002).In both, the user can interact freely with the simulated en-vironment via his or her individual avatar, defined as arepresentation of the user’s identity within the computer en-vironment (Gerhard, 2003; Gerhard, Moore, & Hobbs,2004). Users can typically select the nature of their avatar. Forexample, it might be an entirely abstract shape, a cartoon char-acter (a currently popular avatar, for example, is Homer Simp-son), or a humanlike representation (e.g., a more or lessphoto-realistic head). The user “inhabits” his or her avatar andassumes the viewpoint of the avatar (Cobb et al., 2001).

In an SVE, the user is restricted to interacting with the en-vironment and possibly with other avatars, responses fromwhich currently must be preprogrammed (Cobb et al., 2001),though in the near future they may be driven by more or lessautonomous software “agents” (Gerhard, 2003). In a CVE,there can be more than one user, and users can communicatewith each other via their avatars (see Figure 1). We define aCVE, then, as a distributed computer-based virtual space (or

FOCUS ON AUTISM AND OTHER DEVELOPMENTAL DISABILITIESVOLUME 20, NUMBER 4, WINTER 2005

PAGES 231–243

Collaborative Virtual EnvironmentTechnology for People With Autism

David Moore, Yufang Cheng, Paul McGrath, and Norman J. Powell

at CORNELL UNIV on November 18, 2014foa.sagepub.comDownloaded from


FOCUS ON AUTISM AND OTHER DEVELOPMENTAL DISABILITIES

232

set of spaces) in which people can meet and interact with oth-ers via their avatars (cf. Churchill, Snowdon, & Munro, 2001;Gerhard, 2003). Benford, Bowers, Fahlen, Greenhalgh, andSnowdon (1995) suggested that CVEs have a number of fea-tures: navigation, embodiment, communication, and interac-tion. Similarly, Churchill and Snowdon (1998) pointed outthat a number of key features are supported in CVEs: sharedcontext, awareness of others, negotiation and communication,and flexible and multiple viewpoints. As an example, the Com-munications Research Group (CRG) at the University ofNottingham developed the “MASSIVE” CVE platform (Com-munications Research Group, 1997). This environment allowsseveral users to simultaneously occupy a virtual environmentand interact with other users. It thereby allows individuals sep-arated by distance to participate in virtual meetings. Relatedresearch is “Cyber Axis,” an interactive three-dimensional rep-resentation of an art gallery, which provides a view of art andenables people to discuss, via their avatars, works of art on theInternet (Gerhard, 2003).

The role of the avatar, in providing a visible representativeform of the user, is a crucial element that distinguishes CVEsfrom other virtual social spaces, for example disembodied chatsystems (Rossney, 1996). The avatar is seen as providing afocus for conversation and social interaction (Slater, Sadagic,Usoh, & Schroeder, 2000). Taylor (2002) argued that usersembody themselves through their avatars, which makes theirengagement with the virtual world more real. Avatars may fa-cilitate social encounters in the virtual world and may imply toothers that they are acting with the authority of the underly-ing user. An avatar is a proxy for the purposes of simplifyingand facilitating the process of interhuman communication ina virtual world, as users not only identify with their own avatarsbut also recognize the existence of others through their avatars

(Schroeder, 2002). Given all this, avatars can potentially pro-vide presence and social facilitation for all participants in aCVE (Gerhard, 2003).

A Prima Facie Case for CVE for People With Autism

Autism involves a “triad of impairments” (Wing, 1996). Thereis a social impairment: The person with autism finds it hard torelate to, and empathize with, other people. There is a com-munication impairment: The person with autism finds it hardto understand and use verbal and nonverbal communication.Finally, there is a tendency toward rigidity and inflexibility inthinking, language, and behavior. Some researchers suggestedthat this triad is underpinned by a “theory-of-mind deficit”(e.g., Howlin, Baron-Cohen, & Hadwin, 1999): People withautism may have difficulty understanding mental states inthemselves and others.

Given this understanding of autism, and given the charac-terization of CVEs presented in the previous section, we arguethat CVE technology can potentially benefit people with autismin three ways: as an assistive technology, as an educational tech-nology, and as a means of helping address any theory-of-mind(ToM) impairment.

CVE as an Assistive Technology

Concerning its potential role as an assistive technology, our ar-gument is that people with autism may be able to use the CVEtechnology to communicate more fruitfully with other people(Moore, 1998). The approach here is similar to that of Han-ner et al. (2003), who are studying the use of laptop comput-ers to enable young people with autism to network with eachother. This is important because, as Cobb et al. (2002) noted,people with autism may experience social exclusion becausethey find it difficult to make friends; indeed, Ozonoff andMiller (1995) saw a difficulty relating socially to people as ahallmark of autism. Any means of addressing these issues, weargue, is therefore worthy of investigation.

Further, key aspects of CVE technology suggest that it hasthe potential to be effective in such an assistive technologyrole. The technology enables communication that is simplerand less threatening than its face-to-face equivalent and thatavoids many of its potential pitfalls (Parsons et al., 2000; Par-sons, Mitchell, & Leonard, 2005). Aarons and Gittens (1998)suggested that people with autism have a particular difficultyunderstanding subtle social rules. Because of their remote na-ture, CVEs can enable communication without the need forsuch subtle social cues. Parsons et al. (2000) suggested thatCVEs give users active control over their interactions, and thatthis may increase the confidence of people who otherwise mayfeel out of control in social situations.

These benefits may promote meaningful and interestingcommunication. CVEs facilitate direct communication (Cobb

FIGURE 1. Collaborative virtual environment schematic(after Gerhard, 2003).



VOLUME 20, NUMBER 4, WINTER 2005

233

et al., 2001) and, despite the fact that they are computer-based, represent an unstructured context in which users arefree to make their own choices as they interact with others(Cobb et al., 2002). Parsons and Mitchell (2001) argued thatinteractions via CVE technology tend to be slower than face-to-face interactions, and that slowing down the rate of inter-action may provide users with autism with time to think ofalternative ways of dealing with a particular situation. Further,because of its use of avatars, the technology enables the rep-resentation of emotions, via the avatars’ facial expressions. Thisis important because emotions are needed for good commu-nication (Fabri, Moore, & Hobbs, 1999, 2004) and becauseemotions are conveyed by facial expression better than they areby words (Poggi & Pelachaud, 2000).

Thus, CVE technology can potentially help people withautism who cannot or do not wish to come together physi-cally, but who wish to discuss common interests. The tech-nology may therefore provide a means by which people withautism can communicate with others, and thus circumvent, atleast in part, their social and communication impairment andsense of isolation.

In sum, then, the argument for the use of CVEs as an as-sistive technology is that a CVE can potentially provide ameans by which people with autism can communicate withothers and thus circumvent their social and communicationimpairment and sense of isolation. As Marla Comm has mov-ingly put it, “on the Internet, I found companionship and sup-port in a setting where weak social skills don’t matter” (GlobalIdeas Bank, 2004). We believe that CVE technology can po-tentially enhance this facility.

CVE for Education

Concerning the potential educational use of CVE technology,the idea is to use the technology as a means of educating userswith autism, possibly in an attempt to help overcome theirautism-specific “deficits.” Thus, the interlocutor (conversa-tional partner) of the user with autism in the CVE may be insome sense his or her “teacher.” One specific way in which thismight be used is for the purposes of practice and rehearsal ofevents in the “real world” (e.g., a forthcoming school visit orwedding).

The use of group work in the education of people withautism is often advocated (e.g., Ozonoff & Miller, 1995), andCVE may provide a useful vehicle for such group work, eitherfor geographically remote participants or for a class in a com-puter room conducting a virtual group exercise. Indeed, Nealeet al. (2002) suggested that desktop VE technology is power-ful partly because it can support social interaction around thecomputer. Similarly, social skills programs are also often advo-cated, partly on the grounds that social impairments can affectgeneral educational progress (e.g., Aarons & Gittens, 1998).Preliminary evidence, albeit based on studies involving smallnumbers of participants, suggests that computer-based ap-proaches may form a useful ingredient in such programs (e.g.,

Bernard-Opitz, Sriram, & Nakhoda-Sapuan, 2001; Bishop,2003; Hagiwara & Smith, 1999; Hetzroni & Tannous, 2004).

The argument for the use of VE technology in such pro-grams is that it enables social skills to be practiced and re-hearsed in realistic settings in real time (Cobb et al., 2002;Neale et al., 2002; Parsons, Mitchell, & Leonard, 2004; Par-sons et al., 2005; Trepagnier, 1999). Further, SVE technologyoffers environments that are realistic yet safe and controlledand that can be visited many times (Beardon et al., 2001).Charitos et al. (2000) pointed out that a VE system can re-move distractions that may otherwise have a negative effect onlearners. A range of scenarios for practice can be provided, andlearners can have a secure environment in which they may beable to learn appropriate social rules without having to dealface-to-face with other people (Cobb et al., 2002). Users’ in-teractions within the VE can be recorded and used for subse-quent educational discussion between learner and teacher.

All of these potential advantages of SVEs are inherited byCVEs. In addition, CVEs allow users to practice “real” con-versations in role-play scenarios (Cobb et al., 2002). A CVEcreates the opportunity for people with autism to learn bymaking mistakes but without suffering the real consequenceof their errors. Because they are represented by an avatar, usersmay feel “anonymous” and hence safer.

On the other hand, Jordan (1995), discussing communi-cation programs, warned against teaching communication skillsdivorced from their communicative context. Similarly, Bear-don et al. (2001) argued that people with autism may find itdifficult to apply skills learned in isolation, via VEs, to situa-tions outside the learning environment. Indeed, it might beargued that use of the computer for education may exacerbateany social difficulty of users with autism, causing them to relyon, and perhaps become obsessed with, the computer, andthus engage in less “real” social interaction (cf. Parsons &Mitchell, 2002; Parsons et al., 2000).

There are, we suggest, four answers to such concerns.First, to the extent that the argument made earlier for CVEsas an assistive technology is valid, interactions within a CVEare themselves real—people are engaging in a genuine, albeitnot face-to-face, conversation; they are potentially, therefore,engaging in more, not less, human interaction. Second, VEsare not being advocated as the only approach to education;hence any negative effects can be countered, in principle atleast, by other educational approaches. Third, the concern thatuse of VEs might “collude with” a user’s autism can poten-tially be ameliorated by the adoption of collaborative workingpractices (Parsons & Mitchell, 2002). Finally, Parsons andMitchell (2002) argued that the greater flexibility and unpre-dictability of VEs as compared with conventional computerprograms render them less likely to be used obsessively.

CVE for Theory of Mind

Another interesting possibility is that of using CVE technol-ogy to help children with autism with any ToM deficit. Al-




234

though the status of this alleged deficit is controversial, with,for example, some research suggesting that the perception ofemotions in others is not systematically or specifically deficientin people with autism (cf. Gepner, Deruelle, & Grynfeltt,2001), many advocate its explicit teaching (e.g., Howlin et al.,1999; Ozonoff & Miller, 1995). Howlin et al. (1999) arguedthat children with autism can be successfully taught to inter-pret mental states. They further suggested that a combinationof approaches may turn out to be the most beneficial meansof teaching ToM concepts. Teaching such concepts has beenthe focus of some commercial CBL development work by the University of Cambridge (Baron-Cohen, Hill, Golan, &Wheelwright, 2002) and “Team Asperger” (http://ccoder.com/GainingFace/). Empirical studies have provided someevidence that a CBL approach may be useful (Rajendran &Mitchell, 2000; Silver & Oakes, 2001; Swettenham, 1996), al-though no evidence has thus far been found of successful gen-eralization to participants’ daily lives.

It can be argued that CVEs have a potentially valuable roleto play concerning ToM. Users of a CVE can express theiremotions via choice of an appropriate facial expression for theiravatars (in a similar way to using “emoticons” in chat roomcommunication). Being able to express their emotions, andbeing required to interpret the emotions displayed by their in-terlocutors’ avatars, may help address the ToM issue.

Preliminary observations by Parsons (2001) of people withautism using VEs suggests that they can make appropriate in-ferences concerning what avatars are doing, and we argue thatthe issue of whether they may be able to, or may be able tolearn to, make appropriate inferences concerning the feelingthe avatar represents is worth investigation. McIlhagga andGeorge (1999) suggested that in a CVE, a user sees theavatar’s behavior and facial expressions and hence builds amodel of the emotional state of the underlying agent or user.Enabling people with autism to work in such environmentsprovides them, therefore, in principle at least, with an oppor-tunity to practice their mind reading skills. Gepner et al.(2001) suggested that slow, dynamic presentations may facili-tate recognition of facial expressions by children with autism,and this, together with the ability to adjust the speed of avatarfacial movement, adds to the case for VEs to help with ToM.

Although there is very little evidence specifically concern-ing CVEs and people with autism (cf. the next section of thispaper), it has been argued more generally that CVEs can po-tentially represent everyday human activities and have the abil-ity to support people in using their natural social skills tointeract with others (Benford et al., 1995). Further, Fabri et al. (2004) provided evidence that adults who do not haveautism can successfully recognize emotions as portrayed byavatar representations. Gerhard, Moore, and Hobbs (2001)argued that communication can take place in a CVE withoutthe existence of preconceptions and prejudices based on phys-ical appearance. This general facility can be expected to be “in-herited” when the technology is used by people with autism.

Current Evidence Concerning CVEs and People With Autism

A plausible case can be made, then, for the use of CVE tech-nology by people with autism, and attempts are being madeto investigate its practical efficacy. Charitos et al. (2000) de-veloped a VE system to enable children to rehearse everydayactivities around the home. Similarly, researchers at the Uni-versity of Nottingham have developed an SVE aimed at de-veloping social awareness and social skills in work-relatedscenarios (Beardon et al., 2001) and SVE and CVE versionsof a café (Cobb et al., 2002). Preliminary evidence concern-ing adults with autism suggests that they may have difficultyunderstanding social conventions in a virtual environment(Parsons, 2001). Parsons suggested that this might be becausethey believe that socially inappropriate actions “do not mat-ter” in an environment that is only virtual, or perhaps becausethey perceive operating within the environment as merely a“game” (Parsons, 2001). Similarly, Cobb et al. (2001) foundthat when using a CVE, users could manipulate their avatarsreasonably well but spent very little time communicating witheach other.

On the other hand, Cobb et al. (2002) reported prelimi-nary observations suggesting that users found their virtual en-vironments interesting, fun, and easy to use, and they madeappropriate inferences concerning behavior of avatars. Parsonset al. (2004) studied VE use with a group of 12 participantswith autism (ages 13 to 18), matching them to 12 partici-pants according to verbal IQ and 12 participants according toperformance IQ. They found that the participants with autismwere able to quickly learn how to use the VE equipment. Theyalso found that the majority (10/12) of the participants withautism were able to interpret the VE as being representationalrather than literal, and they argued that this is important, onthe grounds that an unduly literal understanding of a VEwould limit its utility for teaching social skills, as users need tounderstand that what happens in the VE is not necessarily aliteral representation of what would happen in real life. Itmight be argued, though, that this presents a further difficulty,for SVEs if not CVEs, because intuitively the need to under-stand the nonliteral nature of the VE may militate against theease of generalizing lessons from it to the real world. In a re-lated study, Parsons et al. (2005) found that some users of theirVE displayed a limited understanding of it and engaged in con-siderable off-task behavior. They suggested that a mixed mes-sage has emerged from their study regarding the interpretationand understanding of VEs, and they cautioned that VEs maynot be suitable for all people with autism; they also acknowl-edged that the group sizes in the study were small.

Elsewhere, Parsons et al. (2000) argued that the appear-ance on the Internet of chat rooms specifically for people withautism suggests that multiple users can use the same virtual en-vironment effectively. More generally, support for the use ofCVE technology for “special needs” education is provided by




235

Almeida and Ramos (2000), who described a framework forthe use of CVEs to promote the education and social interac-tion of people with Down syndrome.

The work just reported notwithstanding, research con-cerning the use of VEs by people with autism has thus far been very limited (Beardon et al., 2001; Charitos et al., 2000;Cobb et al., 2002; Parsons, 2001; Parsons et al., 2000; Par-sons et al., 2004; Parsons et al., 2005). The work discussed inthis article represents an attempt to contribute to this new re-search area. In sum, there is on the one hand a plausible casefor CVE technology as potentially valuable for people withautism, and on the other hand a lack of evidence regarding itspractical utility. This juxtaposition motivates the research dis-cussed in the next section.

Method

Given the centrality of the avatar to CVEs, our investigationsthus far have concentrated on the ability of people with autismto interact with avatars. This is important, we argue, becauseavatars are central to CVE technology, and because an under-standing of their emotional expressiveness is important toCVE communication in general and to the use of CVEs foraddressing ToM issues in particular. To facilitate such an in-vestigation, we have developed a simple computer system (seeNote; Cheng, 2005). The research interest in the system is tosee it as a context for an empirical study of interactions withavatars. We also hope that the system might be of educationalvalue per se, and have built it with such a dual role (educationand research) in mind.

The System

The system incorporates avatar representations for four emo-tions: happy, sad, angry, and frightened. The precise represen-tations, with corresponding definitions (based on the work ofFabri et al., 2004), are illustrated in Figure 2. The system in-volves three stages, which the user can work through in anyorder. In Stage 1 the avatar representations of the four emo-tions are sequentially presented in isolation, and users areasked to select from a list the emotion they think is being dis-played (see Figure 3). In a second activity in Stage 1, users aretold that a particular emotion is being felt and asked to selectthe avatar head they believe to correspond to that emotion (seeFigure 4); this is repeated for each of the four emotions. Theeducational rationale behind Stage 1 of the system is that itmay be a useful means of helping users recognize basic emo-tions. The research rationale is that an accurate understandingof the avatar emotions is a prerequisite for their successful usewithin a CVE.

Stage 2 attempts to elicit the possible emotions in the con-text of a simple social scenario; it requires users to predict thelikely emotions caused by certain events. Four questions, per-

taining to three emotions, are asked (the “fear” emotion wasdropped from this stage because of a concern that “frighten-ing” social scenarios might be unsettling to users and there-fore be unethical). The questions involve one’s own feelingsand those of others. Figure 5 shows an example of a questionconcerning the emotion the user might feel in a hypotheticalcontext. When the user answers a question, a thought bubbleappears, in an attempt to help the user understand what theperson in question may be feeling. Figure 6 shows the thoughtbubble that appears following a question about the feelings ofthe user’s brother. The use of thought bubbles is designed toprovide reinforcement in the event of a correct answer and ahint in the event of an incorrect answer. The educational ra-tionale behind Stage 2 of the system is based on the educa-tional arguments for Social StoriesTM in general (Gray, 2000)and computer-based social stories in particular (Hagiwara &Smith, 1999). Further, the social scenarios incorporate simplerole plays, in line with Neale’s (2001) argument that the useof role plays is an accepted approach to teaching social skills topeople with autism. Some questions offer more than one cor-rect answer, on the grounds that it is often the case that a givencontext may cause different emotions. The use of thoughtbubbles follows Howlin et al. (1999) and Rajendran andMitchell (2000). The research rationale behind Stage 2 is thatan understanding that events cause emotions is important forCVE use, as users need to understand that events that occurin the CVE (e.g., a message they have sent) may cause emo-tions in other users of the CVE.

In Stage 3 of the system, the user is given an avatar repre-sentation of one of the emotions and asked to select which ofa number of given events he or she thinks may have caused thisemotion (see Figure 7). As with Stage 2, some questions offermore than one correct answer. There are six questions involv-ing three emotions in this stage, inferring from one’s own feel-ings and those of others. The educational rationale behind thisstage is that it rehearses the ability to infer an emotion and itslikely cause, and that this is an important life skill. The researchrationale is that in a CVE, users need to understand that achanged emotion will have been caused by some event or sit-uation, rather than being a random occurrence.

The avatar faces use simple animated sequences, in linewith claims that a still display does not capture the realistic na-ture of facial expressions that will be encountered in everydaysituations (Gepner et al., 2001; Trepagnier, 1999). The use ofanimation is also in line with Silver and Oakes’s (2001) pointthat children with learning disabilities prefer programs thatinclude animation, sound, and voice. This was further accom-plished by having text in the system read out (via computer-generated speech), an aspect that was also designed to obviatethe need for users to have good reading skills to be able to usethe system. The system was kept as simple as possible, consis-tent with its aims, in case of possible sensory dysfunction inthe intended user population (Aarons & Gittens, 1998). Com-ments on an earlier version of the system from an informal



236

Happy

• The brow is relaxed. There is no distinctive brow or forehead movement in happiness.

• The lower eyelid may show wrinkles below it and may be raised. Wrinkles can also go outwardfrom the outer corners of the eyes.

• The mouth is widened in front and pulled back towards the ears. The cheeks are raised, andwrinkle runs down from the nose to the outer edge beyond the lip corners.

Anger is manifested in each of the three facial areas:

• The inner portion of eyebrows is draw down toward the nose. There is a frown on the brow.

• The upper and lower eyelids are tense. But the upper lid may or not be lowered by the action ofthe brow. Likewise, the lower lid may or may not be raised. The eyes themselves stare out in apenetrating fashion.

• The mouth is commonly compressed, and lips are pressed firmly together with corners straight.

• The brows are drawn together and the inner portion is pulled upward. The inner corner of theupper eyelid is raised.

• The mouth is closed and the corners are pulled in to a frown. The lower lip is curled inward,blending into the chin.

• The eyebrows are lifted straight up and drawn together in the middle, so that the inner cornersare closer together.

• The eyes are opened widely. The upper eyelid is raised and the lower eyelid tensed and drawnup.

• The mouth is open and the lips are either tensed or stretched and may be drawn back tightly.The lower lip is moved upward and outward.

• The jaw is open slightly.

Angry

Sad

Frightened

FIGURE 2. Specification of the avatar representations of emotions.




237

focus group comprising parents of children with autism wereincorporated into the final design.

Procedure

A copy of the SVE system as just outlined was mailed to 100potential participants. Participants’ mailing addresses were ob-tained via requests published on the Internet and sent to vari-ous autism e-mailing lists and schools in the United Kingdom;some of the schools specifically serve students with autism,others serve both children who have and children who do nothave autism. From responses to this, packs were sent to theparents, teachers, or guardians of 100 potential participants.Potential participants who had not replied within a monthwere contacted by telephone or e-mail. From the 100 poten-tial participants to whom packs were sent, 34 replied, for a re-sponse rate of 34%. The age range of the participants was from7.8 to 16 years, with a mean age of 9.96. Twenty-nine of theparticipants were boys and 5 were girls. All resided in theUnited Kingdom. Participants were asked to work through thethree stages of the SVE system described above. The softwarelogged their work onto a diskette that had been sent in theirpack. After the users had operated the software, the users andtheir parents were each asked to fill in a questionnaire. Usersthen returned both the disk with the log data and their com-pleted questionnaires.

Data Analysis and Results

To establish whether the participants are successfully identify-ing the emotional avatars, we need to consider first what itwould mean if they were not. In this case, the participantswould be responding to the questions, effectively unaware ofthe meaning of the expressions, relying on chance and guess-work to make their choices. Our method of data analysis,therefore, compared the observed responses of the participantsto the questions against the responses that would be expectedwere they to be selected by chance.

The SVE contains a total of 18 questions of three differ-ent types: questions for which there was one correct answerfrom three options, questions with one correct answer fromfour options, and questions with two correct answers fromthree options. For each type of question, there is a probabilityassociated with the appropriate answer being chosen by chanceon the first attempt. The average number of questions ex-pected to be answered correctly on the first attempt, N, isgiven by

N = Σ nipi = 9 × 3 + 4 × 4 = 5 × 3 = 3 = 7.3

where ni is the question number and pi is the chance of an-swering that question correctly on the first attempt by chance(the dots above the digits represent recurring numbers, so 7.3 above represents 71⁄3 and 0.407 [below] represents0.407407407…).

211 22

FIGURE 3. Stage 1 of the system.

FIGURE 4. Example of the second activity in Stage 1 of thesystem.


.

i

. . .




238

Table 1 shows the number of each participant’s correct an-swers to the 18 questions and that participant’s significancelevel when compared with what would be expected by chance.A one-tailed significance test, against the exact version of thebinomial distribution with a proportion of 0.407 (7.3/18)correct answers on the first attempt, was employed. Table 1shows that 30 out of the 34 participants answered the ques-tions at a level significantly better than would be expected bychance (p < .05).

Figure 8 contains a histogram of the number of correct an-swers on the first attempt to all the questions, with the num-ber expected by chance indicated by the vertical reference line.As can be seen, the distribution is grouped toward the maxi-mum number of questions, with the majority of participantsscoring well above the values expected by chance. The sign testconfirms that the result is very highly significantly different (31above, 3 below, and 0 ties, Z = 4.63, p < .0005) from the valuethat would be expected by chance. (The sign test was used be-

cause it makes no assumptions about the underlying distribu-tion of the data, which in the current case is very left skewed.)

Overall, the analysis indicates a high level of success at an-swering the questions and (by the same token) recognizing theemotions being portrayed by the avatars; 30 out of the 34 par-ticipants were able to use the avatars at levels demonstrablybetter than chance. Thus, in general, the study offers evidencethat the participants who responded understood the emotionsof the avatars appropriately.

Discussion

In this article, we have made a prima facie case for CVE tech-nology as potentially valuable for people with autism, and wehave discussed an exploratory empirical study we conductedconcerning an important ingredient in successful use of CVEs,

FIGURE 6. Thought bubble used in Stage 2 of the system.


TABLE 1Analysis of Participants’ Answers

Number of questions Probability (p) of thiscorrectly answered occurring by chance

Participant on first attempt (exact binomial test)

1 13 0.0072 6 0.3513 14 0.0024 8 0.4615 17 < 0.00056 17 < 0.00057 17 < 0.00058 17 < 0.00059 14 0.002

10 18 < 0.000511 16 < 0.000512 16 < 0.000513 17 < 0.000514 16 < 0.000515 17 < 0.000516 16 < 0.000517 16 < 0.000518 14 0.00219 13 0.00720 14 0.00221 18 < 0.000522 18 < 0.000523 16 < 0.000524 15 < 0.000525 15 < 0.000526 17 < 0.000527 6 0.35128 17 < 0.000529 15 < 0.000530 6 0.35131 17 < 0.000532 17 < 0.000533 14 0.00234 12 0.024

. . .




239

namely, the ability to recognize the emotional expressions in-tended to be portrayed by the avatars used in the CVE.

The study demonstrated an ability of the majority of theparticipating group to correctly interpret the avatars’ repre-sentations of emotions to a level where they could perform avariety of tasks. These tasks involved recognition of an emo-tion from an expression, selection of an expression to repre-sent an emotion, prediction of an expression from a simplesocial scenario, and the inference of a causal event from an ex-pression. The prediction and inference tasks were expressedfrom both the participant’s and another person’s point of view.While these tasks have not been tested extensively, in that onlya single simple social scenario has thus far been explored, thestudy demonstrated ability at all these tasks.

We argue that these tasks are all prerequisites to function-ing effectively in a CVE with expressive avatars and thatdemonstrating ability at these tasks is therefore an importantstep in demonstrating the feasibility of these CVE systems foruse with children with autism. The technology itself is verynew, and the notion of using it to help people with autism isnewer still. As a result, our study has, perforce, been of an ex-ploratory nature. Nevertheless, the study does, we contend,give grounds for optimism that the potential advantages ofCVEs argued for earlier can ultimately be achieved.

Throughout the system used in the study, the avatar “face”is used as the means of portraying emotions. This is in line withresearch suggesting that facial expressions are the primarymeans by which children read emotions (George & McIl-hagga, 2000; McIlhagga & George, 1999) and that childrenwith autism tend to be impaired in their ability to process fa-

cial expressions (e.g., Gepner et al., 2001). In line with relatedresearch concerning avatar understanding by adults who donot have autism (e.g., Gerhard, 2003), we use a humanoidavatar at all times. Similarly, because the system is aimed pri-marily at children, we use avatar representations based onyoung people, in line with evidence that children are better atrecognition of faces when given similarly aged faces to con-sider (George & McIlhagga, 2000).

A more problematic issue when we were developing thesystem concerned the range of emotions to consider. The lit-erature suggested that there are seven “universal” emotions(Ekman, 1982), and research with adults who do not haveautism has indicated that they can be successfully representedvia avatar faces (Fabri et al., 2004). George and McIlhagga(2000), however, argued that it is debatable when and if chil-dren utilize all seven, and work with individuals with autismtends to concentrate on a subset of emotions. The precise con-tent of this subset tends to vary, however. Our decision to usehappy, sad, angry, and frightened follows Howlin et al. (1999).

Another interesting issue concerns the “realism” of theavatars. On the one hand, there is evidence that children findcaricature faces easier to understand than normal faces (e.g.,George & McIlhagga, 2000). On the other hand, use of morerealistic faces is intuitively more likely to prove helpful whenchildren are operating in the real world. Our approach herewas to base the avatar representations on the work of Fabri(e.g., Fabri et al., 2004). Basing the representations on Fabri’swork has two advantages. First, there is evidence, albeit notfrom people with autism, that the representations are readilyunderstandable. Second, Fabri’s representations are them-

Fre

quen

cy

FIGURE 8. Histogram of the number of correct first attempts.




240

selves derived from accepted standards for the representationof emotions (see Fabri et al., 2004, for a detailed discussion ofthis point), and basing the representations on accepted stan-dards increases one’s confidence in them. Further, by study-ing the standard representations as they are used by individualswith autism, we are arguably testing the standard in extremis,and hence potentially enabling the standard to be strength-ened. This is an example of the “offshoot” argument for as-sistive technology: Lessons from the use of the technology inextraordinary human computer interaction might lead to help-ful development of the technology for “general” use (cf. Ed-wards, 1995; Hobbs & Moore, 1998). Similarly, George andMcIlhagga (2000) pointed to a lack of guidance in the litera-ture as to how children might understand the behavior of vir-tual characters and their emotional signals, and our work canbe expected to contribute here.

Given the positive results of our exploratory study and thefact that resources specifically designed to help address socialskills difficulties for people with autism remain limited (Neale,2001), CVE technology should continue to be investigated asa potential ingredient in such resources. There are many waysin which such an investigation could usefully be pursued. First,a bigger and more controlled study using the current com-puter system could be carried out. In particular, there is evi-dence from our study that some participants, albeit a smallminority, had a real difficulty in understanding the emotionalrepresentation of the avatars. The needs of such participantsrequire further study. It may be, for example, that in any de-velopment of CVE systems, this minority should be recog-nized and catered to, either through additional training orextra cues. The nature and degree of their difficulty and theform and effectiveness of any extra support provided will allrequire further investigation.

Second, there are many changes that could be made to thecurrent system, the effectiveness of which would bear investi-gation. It would be interesting to follow George and McIl-hagga (2000), using scales of emotion (e.g., happy–sad) ratherthan discrete labels and varying the emotion displayed by theavatar face. The system currently has no “neutral” avatar face,and all avatars are young, White, and male; changes to theseparameters could be experimented with. Similarly, a greaterrange of emotions, and changing the timings of the animatedtransitions between representations of different emotions,could be studied. Again, rather than concentrating on mere fa-cial representations, full body animations, involving potentialemotion signal carriers such as posture (cf. Fabri et al., 2004)could be implemented. More ambitiously, perhaps, the systemcould be made to act as a “shell,” allowing teachers or studentsto configure the way in which the avatars represent the vari-ous emotions. It might well be, for example, that users of thesystem improve their performance with it through time, andhence the cues can be amended to become more subtle (McIl-hagga & George, 1999).

A third important area of further work is to investigate em-pirically some of the arguments we presented earlier as part of

the case for CVE technology. In particular, we have tried toargue that in principle people could use the technology to in-crease their social interactions, by, for example, communicat-ing via the CVE with people (who may or may not haveautism) on topics of mutual interest. Such communicationsmight be impossible in the absence of the technology, eitherbecause the interlocutors are geographically remote or becauseof a possible reluctance of the person with autism to engagein face-to-face conversation. The promise of the technology,then, is greater social interaction, albeit not face-to-face com-munication of the hitherto conventional sort. The extent towhich this might happen in practice, however, is an empiricalquestion, well worthy of future investigation. Similarly, the ex-tent to which (if at all) social skills practiced within a CVE gen-eralize to social skills in face-to-face situations also needsempirical investigation (cf. Parsons et al., 2005). Again, the ar-gument of Parsons and Mitchell (2002), cited earlier, that VEtechnology is less likely than other forms of computer tech-nology to be used in an obsessional manner, also merits em-pirical study.

Our immediate next step, however, is to simulate a morenaturalistic CVE. It is important to observe whether the abil-ities to recognize avatar emotion representations discussedabove are also displayed in a more realistic collaborative virtualenvironment, involving interaction with others and hence op-portunity and purpose for emotional expression. It will be in-structive to note whether in this environment emotionalexpressions will be employed, registered, and responded to.We have conducted an initial investigation of this and will re-port the results in a future article.

Certain methodological considerations should be borne inmind when considering the results presented in the previoussection. It might be argued that the study is relatively lackingin control, because the manner in which participants engagedin the exercises making up the study was largely left to theirdiscretion, given that the exercises were mailed to them. Onthe other hand, and by the same token, because the childrenwere engaging in the study in their usual everyday environ-ments, and because this is the environment in which CVEtechnology would eventually be used, the study context is ar-guably a realistic one. There is often a trade-off between con-trol and realism (cf. Moore, Hobbs, Mullier, & Bell, 1997),and our current study has favored the latter, albeit perhaps atthe expense of the former.

A specific control-related issue is that the study relied onthe parents or teachers of participants to report the autism di-agnosis of the specific participant in their care, rather thanadministering an autism diagnosis tool. However, the partici-pants were drawn from schools serving students with autismand from autism e-mailing lists, and this reliance on self-reportis therefore perhaps a minor concern, especially given the lo-gistic difficulties and ethical issues that would be associatedwith administering a diagnostic test to each participant. How-ever, it does suggest that caution must be taken when consid-ering the results.




241

Another reason for caution is that the overall sample sizeof 34 is small. In response to this, we point out that the sam-ple size is bigger than in many studies of CBL technologies forpeople with autism. More positively, the statistical analyses takeaccount of the sample size (Cheng, 2005). Caution is still re-quired, however. The response rate to the study was only 34%,and there may be a link between nonresponse and perfor-mance. For example, people who would find the task more dif-ficult may be less likely to respond; conversely, it may be thatthe young people were helped in the tasks by their parents orcaregivers; either way, the reported success rates of avatar usecould be overly optimistic.

These methodological issues notwithstanding, we contendthat the study provides evidence that the majority of childrenwith autism in the age range studied will be able to use emo-tional avatars. The results add to the case for the suitability ofthe standard for avatar representations that was adopted forthe study. Similarly, George and McIlhagga (2000) pointed toopen questions concerning the design of affective behavior ofvirtual characters for children, and there is much recent inter-est in the field of “affective computing” in general (see, e.g.,Hudlicka, 2003; Picard, 2003); our results may contribute tothis research endeavor. Most important, however, the resultsgive grounds for optimism that the potential benefits of CVEtechnology argued for earlier may be realized, and they sug-gest therefore that CVEs are worthy of further research as apotentially valuable technology for people with autism.

ABOUT THE AUTHORS

David Moore, PhD, is a principal lecturer and teacher fellow in theSchool of Computing, Leeds Metropolitan University, United Kingdom.His research interests are computer systems for people with autism, vir-tual environments, and computational dialectics. Yufang Cheng, PhD,is an assistant professor in the Department of Business Education of National Changhua University of Education in Taiwan. Her currentinterests are digital media development, digital media learning con-cerning disabilities, and interactive multimedia applications. Paul Mc-Grath, BSc(Hons), is associate dean and head of school of computing atLeeds Metropolitan University. His current interests include interfacedesign methods, software process improvement, and education for pro-fessional computing practice. Norman J. Powell, PhD, is a problem-based learning officer, Manchester University. His current interestsinclude learning disabilities, especially dyslexia and autism, and theteaching and learning of computer programming and electronic engi-neering. Address: David Moore, School of Computing, Leeds Metropoli-tan University, Beckett Park, Leeds LS6 3QS, UK; e-mail: [email protected]

AUTHORS’ NOTE

A copy of the system may be obtained by from David Moore (e-mail:[email protected]).

REFERENCES

Aarons, M., & Gittens, T. (1998). Autism: A social skills approach forchildren and adolescents. Oxford, UK: Winslow Press.

Almeida, A. M. P., & Ramos, F. M. S. (2000). Collaborative net-worked framework for the rehabilitation of children with Down’ssyndrome. In P. Sharkey, A. Cesarani, L. Pugnetti, & A. Rizzo(Eds.), Proceedings of the 3rd International Conference on Disabil-ity, Virtual Reality and Associated Technologies (pp. 281–286).Reading, UK: University of Reading.

Baron-Cohen, S., Hill, J., Golan, O., & Wheelwright, S. (2002).Emotions: An interactive electronic guide. Communication, 36(3),31–32.

Beardon, L., Parsons, S., & Neale, H. (2001). An interdisciplinaryapproach to investigating the use of virtual reality environmentsfor people with Asperger syndrome. Educational and Child Psy-chology, 18(2), 53–62.

Benford, S., Bowers, J., Fahlen, L., Greenhalgh, C., & Snowdon, D.(1995). User embodiments in collaborative virtual environments.In Proceedings of the ACM Conference on Human Factors in Com-puting Systems (CHI’95; pp. 242–249). New York: Association forComputing Machinery.

Bernard-Opitz, V., Sriram, N., & Nakhoda-Sapuan, S. (2001). En-hancing social problem solving in children with autism and normalchildren through computer-assisted instruction. Journal of Autismand Developmental Disorders, 31, 377–384.

Bishop, J. (2003). The internet for educating individuals with socialimpairments. Journal of Computer Assisted Learning, 19, 546–556.

Charitos, D., Karadanos, G., Sereti, E., Triantafillou, S., Koukouvi-nou, S., & Martakos, D. (2000). Employing virtual reality for aid-ing the organisation of autistic children’s behaviour in everydaytasks. In P. Sharkey, A. Cesarani, L. Pugnetti, & A. Rizzo (Eds.),Proceedings of the 3rd International Conference on Disability, Vir-tual Reality and Associated Technologies (pp. 147–152). Reading,UK: University of Reading.

Cheng, Y. (2005). An avatar representation of emotion in collabora-tive virtual environment (CVE) technology for people with autism.Doctoral dissertation, Leeds Metropolitan University, Leeds, UK.

Churchill, E. F., & Snowdon, D. (1998). Collaborative virtual envi-ronments: An introductory review of issues and systems. VirtualReality, 3(1), 3–15.

Churchill, E. F., Snowdon, D., & Munro, A. (2001). Collaborativevirtual environments: Digital places and spaces for interaction. Lon-don, UK: Springer Verlag.

Cobb, S., Beardon, L., Eastgate, R., Glover, T., Kerr, S., Neale, H.,et al. (2002). Applied virtual environments to support learning ofsocial interaction skills in users with Asperger’s Syndrome. DigitalCreativity, 13(1), 11–22.

Cobb, S., Kerr, S., & Glover, T. (2001). The AS Interactive Project:Developing virtual environments for social skills training in userswith Asperger syndrome. In K. Dautenhahn (Ed.), Robotic and vir-tual interactive systems in autism therapy (Communications of theAdaptive Systems Research Group, University of Hertfordshire, Re-port No. 364). Hatfield, UK: University of Hertfordshire.

Communications Research Group. (1997). The CRG virtual envi-ronment (CVE). Retrieved January 19, 2004, from http://www.crg.cs.nott.ac.uk/research/systems/MASSIVE-2/#publications

Edwards, A. D. N. (Ed.). (1995). Extra-ordinary human–computerinteraction. New York: Cambridge University Press.

Ekman, P. (1982). Emotion in the human face. New York: CambridgeUniversity Press.

Fabri, M., Moore, D. J., & Hobbs, D. J. (1999). The emotionalavatar: Nonverbal communication between inhabitants of collabo-rative virtual environments. In A. Braffort, R. Gherbi, S. Gibet, J.




242

Richardson, & D. Teil (Eds.), Gesture-based communication inhuman–computer interaction (Springer lecture notes in artificial in-telligence 1739; pp. 245–248). Berlin: Springer.

Fabri, M., Moore, D. J., & Hobbs, D. J. (2004). Mediating the ex-pression of emotion in educational collaborative virtual environ-ments: An experimental study. International Journal of VirtualReality, 7(2), 66–81.

George, P., & McIlhagga, M. (2000). The communication of mean-ingful emotional information for children interacting with virtualactors. In A. M. Paiva (Ed.), Affective interactions: Towards a newgeneration of computer interfaces (Lecture notes in artificial intelli-gence 1814; pp. 35–48). Berlin: Springer-Verlag.

Gepner, B., Deruelle, C., & Grynfeltt, S. (2001). Motion and emo-tion: A novel approach to the study of face processing by youngautistic children. Journal of Autism and Developmental Disorders,31, 37–45.

Gerhard, M. (2003). A hybrid avatar/agent model for educational col-laborative virtual environments. Unpublished doctoral dissertation,Leeds Metropolitan University, Leeds, UK.

Gerhard, M., Moore, D. J., & Hobbs, D. J. (2001). Continuouspresence in collaborative virtual environments: Towards the evalu-ation of a hybrid avatar–agent model for user representation. In A. de Antonio, R. Aylett, & D. Ballin (Eds.), Proceedings of theInternational Conference on Intelligent Virtual Agents (IVA01; pp. 137–155). Madrid, Spain.

Gerhard, M., Moore, D. J., & Hobbs, D. (2004). Embodiment andcopresence in collaborative interfaces. International Journal ofHuman–Computer Studies, 61(4), 453–480.

Global Ideas Bank. (2004). Computer jobs for those with autism. Re-trieved September 29, 2004, from http://www.globalideasbank.org/site/bank/idea.php?ideaId=288

Gray, C. (2000). The new social story book: Illustrated edition. Arling-ton, TX: Future Horizons.

Hagiwara, T., & Smith, M. B. (1999). A multimedia social story in-tervention: Teaching skills to children with autism. Focus onAutism and Other Developmental Disabilities, 14, 82–95.

Hanner, S., Tomkinson, C., Byrne, J., Boadle, H., Lashley, N., &Stampone, S. (2003). Results after one year of facilitating autismfriendly socialisation (FAFS) using laptops to network young peo-ple with autistic spectrum disorder [Abstract]. In Lisboa 2003— International-Autism-Europe Congress (p. P31). Lisbon, Portugal:Autism-Europe.

Hardy, C., Ogden, J., Newman, J., & Cooper, S. (2002) Autism andICT: A guide for teachers and parents. London: David Fulton.

Hetzroni, O. E., & Tannous, J. (2004). Effects of a computer-basedintervention program on the communicative functions of childrenwith autism. Journal of Autism and Developmental Disorders, 34,95–113.

Hobbs, D. J., & Moore, D. J. (1998). Human computer interaction.London: FTK Publishing.

Howlin, P., Baron-Cohen, S., & Hadwin, J. (1999). Teaching chil-dren with autism to mind-read: A practical guide for teachers andparents. West Sussex, England: Wiley.

Hudlicka, E. (2003). To feel or not to feel: The role of affect inhuman–computer interaction. International Journal of Human–Computer Studies, 59, 1–32.

Jordan, R. (1995, January). Communicating with a computer: The useand abuse of computers in developing communication skills in indi-viduals with autism. Paper presented to the Autisme-France 3rd In-

ternational Conference Autism and Computer Applications, Nice,France.

McIlhagga, M., & George, P. A. (1999). Communicating meaning-ful emotional information in a virtual world. In A. Paiva & C. Mar-tinho (Eds.), Proceedings of the Workshop on Affect in Interactions—Towards a new generation of interfaces (pp. 150–155). Siena, Italy:i3net.

Moore, D. J. (1998). Computers and people with autism/AspergerSyndrome. Communication, Autumn, 20–21. Moore, D. J. (2005). Recent work in computer systems for people with

autism (Internal report). Leeds, UK: Leeds Metropolitan Univer-sity, Leeds.

Moore, D. J., Hobbs, D. J., Mullier, D., & Bell, C. (1997) Interac-tion paradigms with educational hypermedia. In EuroMicro 97—Proceedings of the 23rd EUROMICRO Conference (pp. 65–71). LosAlamitos, CA: IEEE Computer Society.

Moore, D. J., McGrath, P., & Thorpe, J. (2000). Computer aidedlearning for people with autism: A framework for research and de-velopment. Innovations in Education and Training International,37, 218–228.

Moore, D. J., & Taylor, J. (2000). Interactive multimedia systemsfor people with autism. Journal of Educational Media, 25, 169–177.

Neale, H. (2001). Guidance to support the design of virtual envi-ronments to teach social skills. In K. Dautenhahn (Ed.), Roboticand virtual interactive systems in autism therapy (Communicationsof the Adaptive Systems Research Group, University of Hertford-shire, Report No. 364). Hatfield, UK: University of Hertfordshire.

Neale, H., Cobb, S., & Wilson, J. (2002). A front-ended approachto the user-centred design of virtual environments. IEEE VirtualReality 2002: Proceedings (pp. 191–198). Los Alamitos, CA: IEEE.

Ozonoff, S., & Miller, J. N. (1995). Teaching theory of mind: A newapproach to social skills training for individuals with autism. Jour-nal of Autism and Developmental Disorders, 25, 415–433.

Parsons, S. (2001). Social conventions in virtual environments: In-vestigating understanding of personal space amongst people withautistic spectrum disorders. In K. Dautenhahn (Ed.), Robotic andvirtual interactive systems in autism therapy (Communications ofthe Adaptive Systems Research Group, University of Hertfordshire,Report No. 364). Hatfield, UK: University of Hertfordshire.

Parsons, S., Beardon, L., Neale, H., Reynard, G., Eastgate, R., Wil-son, J. R., et al. (2000). Development of social skills amongst adultswith Asperger’s syndrome using virtual environments: The “ASInteractive Project.” In P. Sharkey, A. Cesarani, L. Pugnetti, & A. Rizzo (Eds.), Proceedings of the 3rd International Conference onDisability, Virtual Reality and Associated Technologies (pp. 163–170). Reading, UK: University of Reading.

Parsons, S., & Mitchell, P. (2001). Virtual reality and people withautistic spectrum disorders: Informal observations from the AS In-teractive Project. Proceedings of the 12th Annual Durham Interna-tional Conference on Autism. Sunderland, UK: Autism ResearchUnit/Autism North.

Parsons, S., & Mitchell, P. (2002). The potential of virtual reality insocial skills training for people with autistic spectrum disorders.Journal of Intellectual Disability Research, 46, 430–443.

Parsons, S., Mitchell, P., & Leonard, A. (2004). The use and under-standing of virtual environments by adolescents with autistic spec-trum disorders. Journal of Autism and Developmental Disorders, 34,449–466.




243

Parsons, S., Mitchell, P., & Leonard, A. (2005). Do adolescents withautistic spectrum disorders adhere to social conventions in virtualenvironments? Autism, 9, 95–117.

Picard, R. W. (2003). Affective computing: Challenges. Interna-tional Journal of Human–Computer Studies, 59, 55–64.

Poggi, I., & Pelachaud, C. (2000). Performative facial expressions inanimated faces. In J. Cassell, J. Sullivan, S. Prevost, & E. Churchill(Eds.), Embodied conversational agents (pp. 155–188). Cambridge,MA: MIT Press.

Rajendran, G., & Mitchell, P. (2000). Computer mediated interac-tion in Asperger’s syndrome: The Bubble Dialogue program. Com-puters & Education, 35, 187–207.

Rossney, R. (1996). Metaworlds. Wired, 4(6), 140–152.

Schroeder, R. (2002). Social interaction in virtual environments: Key issues, common themes, and a framework for research. In R. Schroeder (Ed.), The social life of avatars (pp. 1–18). London:Springer.

Silver, M., & Oakes, P. (2001). Evaluation of a new computer inter-vention to teach people with autism or Asperger Syndrome torecognise and predict emotions in others. Autism, 5(3), 299–316.

Slater, M., Sadagic, A., Usoh, M., & Schroeder, R. (2000). Small-group behaviour in a virtual and real environment: A comparativestudy. Presence. Teleoperators and Virtual Environments, 9(1), 37–51.

Swettenham, J. (1996). Can children with autism be taught to un-derstand false belief using computers? Journal of Child Psychologyand Psychiatry, and Allied Disciplines, 37, 157–165.

Taylor, T. L. (2002). Living digitally: Embodiment in virtual worlds.In R. Schroeder (Ed.), The social life of avatars (pp. 40–62). Lon-don: Springer.

Trepagnier, C. G. (1999). Virtual environments for the investigationand rehabilitation of cognitive and perceptual impairments. Neuro-Rehabilitation, 12, 63–72.

Wing, L. (1996). The autistic spectrum: A guide for parents and pro-fessionals. London: Constable.



Documents

Collaborative Virtual Environment Technology for People With Autism