The Effect of Affect: Modeling the Impact of Emotional State on the Behavior of Interactive Virtual Humans Stacy Marsella, Jonathan Gratch, and Jeff Rickel 1 2 1 USCInformationSciencesInstitute 1 4676AdmiraltyWay,Suite1001,MarinadelRey,CA,90292 [email protected],[email protected] USCInstituteforCreativeTechnologies 2 13274FijiWay,Suite600,MarinadelRey,CA,90292 [email protected] ABSTRACT gagingscenariothroughwhichtheywillgainvaluableexperience. A person’s behavior provides significant information about their Forexample,ayoungArmylieutenantcouldbetrainedforapeace- emotionalstate,attitudes,andattention. Ourgoalistocreatevir- keeping missionbyputtinghiminvirtualBosniaandpresenting tualhumansthatconveysuchinformationtopeoplewhileinteract- himwiththesortsofsituationsanddilemmasheislikelytoface. ingwiththeminvirtualworlds.Thevirtualhumansmustrespond Insuchscenarios,virtualhumanscanplayavarietyofroles,such dynamicallytotheeventssurroundingthem,whicharefundamen- as an experienced sergeant serving as a mentor, soldiers serving tallyinfluenced by users’ actions, whileproviding an illusionof ashisteammates,andthelocalpopulace. Unlessthelieutenantis human-likebehavior.Ausermustbeabletointerpretthedynamic trulydrawnintothescenario,hisactionsareunlikelytoreflectthe cognitiveandemotionalstateofthevirtualhumansusingthesame decisionshewillmakeunderstressinreallife. Theeffectiveness nonverbalcuesthatpeopleusetounderstandoneanother.Towards of thetrainingdepends on our success increatingengaging, be- these goals, we are integrating and extending components from lievablecharactersthatconveyarichinnerdynamicsthatunfolds threepriorsystems: avirtualhumanarchitecturewitharangeof inresponsetothescenario. cognitiveandmotorcapabilities,amodelofemotionalappraisal, andamodeloftheimpactofemotionalstateonphysicalbehavior. Thus,ourdesignofthevirtualhumansmustsatisfythreerequire- Wedescribethekeyresearchissues,ourapproach, andaninitial ments. First, theymustbebelievable; thatis, theymustprovide implementationinanArmypeacekeepingscenario. a sufficient illusion of human-like behavior that the human user will be drawn into the scenario. Second, they must be respon- sive; that is, they must respond to the events surrounding them, 1. INTRODUCTION whichwillbefundamentallyinfluencedbytheuser’sactions. Fi- Aperson’semotionalstateinfluencestheminmanyways. Itim- nally,theymustbeinterpretable;theusermustbeabletointerpret pactstheirdecisionmaking,actions,memory,attention,voluntary theirresponsetosituations,includingtheirdynamiccognitiveand muscles, etc., all of which may subsequently impact their emo- emotional state, using the same nonverbal cues that people use tional state (e.g., see [4]). This pervasive impact is reflected in tounderstandoneanother. Thus,ourvirtualhumanscannotsim- thefactthatapersonwillexhibitawidevarietyofnonverbalbe- plycreateanillusionoflifethroughcleverlydesignedrandomness haviors consistent with their emotional state, behaviors that can intheirbehavior;theirinnerbehaviormustrespondappropriately serve a variety of functions both for the person exhibiting them toadynamically unfolding scenario, andtheir outward behavior aswellasforpeopleobservingthem. Forexample,shakingafist mustconveythatinnerbehavioraccuratelyandclearly. atsomeoneplaysanintendedroleincommunicatinginformation. Ontheotherhand,behaviorssuchasrubbingone’sthigh,averting Thispaperdescribesourprogresstowardsamodeloftheoutward gaze,orafacialexpressionoffearmayhavenoexplicitlyintended manifestationsofanagent’semotionalstate. Ourworkintegrates roleincommunication. Nevertheless, theseactions maysuggest three previously implemented systems. The first, Steve [18, 20, considerableinformationaboutaperson’semotionalarousal,their 19], providesanarchitectureforvirtualhumansin3Dsimulated attitudes,andtheirfocusofattention. worlds that can interact with human users as mentors and team- mates. Although Steve did not include any emotions, his broad Ourgoalistocreatevirtualhumansthatconveythesetypesofin- capabilities provide afoundation for thevirtual humans towards formationtohumanswhileinteractingwiththeminvirtualworlds. whichweareworking. Thesecond, E´mile[8], focusesonemo- Weareinterestedinvirtualworldsthatofferhumanusersanen- tionalappraisal:howemotionsarisefromtherelationshipbetween environmental events and an agent’splans and goals. Thethird, IPD [11], contributes a complementary model of emotional ap- praisal as well as a model of the impact of emotional state on physicalbehavior.Theintegrationofthesethreesystemsprovides aninitialmodelofvirtualhumansforexperientiallearninginen- gagingvirtualworlds. Ourworkispartofalargerefforttoadda varietyofnewcapabilitiestoSteve,includingmoresophisticated support for spoken dialog and a more human-like model of per- ception[17]. Report Documentation Page Form Approved OMB No. 0704-0188 Public reporting burden for the collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources, gathering and maintaining the data needed, and completing and reviewing the collection of information. Send comments regarding this burden estimate or any other aspect of this collection of information, including suggestions for reducing this burden, to Washington Headquarters Services, Directorate for Information Operations and Reports, 1215 Jefferson Davis Highway, Suite 1204, Arlington VA 22202-4302. Respondents should be aware that notwithstanding any other provision of law, no person shall be subject to a penalty for failing to comply with a collection of information if it does not display a currently valid OMB control number. 1. REPORT DATE 3. DATES COVERED 2001 2. REPORT TYPE 00-00-2001 to 00-00-2001 4. TITLE AND SUBTITLE 5a. CONTRACT NUMBER The Effect of Affect: Modeling the Impact of Emotional State on the 5b. GRANT NUMBER Behavior of Interactive Virtual Humans 5c. PROGRAM ELEMENT NUMBER 6. AUTHOR(S) 5d. PROJECT NUMBER 5e. TASK NUMBER 5f. WORK UNIT NUMBER 7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) 8. PERFORMING ORGANIZATION University of California,Information Sciences Institute ,4676 Admiralty REPORT NUMBER Way,Marina del Rey,CA,90292 9. SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES) 10. SPONSOR/MONITOR’S ACRONYM(S) 11. SPONSOR/MONITOR’S REPORT NUMBER(S) 12. DISTRIBUTION/AVAILABILITY STATEMENT Approved for public release; distribution unlimited 13. SUPPLEMENTARY NOTES 14. ABSTRACT 15. SUBJECT TERMS 16. SECURITY CLASSIFICATION OF: 17. LIMITATION OF 18. NUMBER 19a. NAME OF ABSTRACT OF PAGES RESPONSIBLE PERSON a. REPORT b. ABSTRACT c. THIS PAGE 5 unclassified unclassified unclassified Standard Form 298 (Rev. 8-98) Prescribed by ANSI Std Z39-18 Figure1:Aninteractivepeacekeepingscenariofeaturing(lefttoright)asergeant,amother,andamedic 2. EXAMPLE SCENARIO or appraisal driven. Communication driven means that one se- To illustrateour vision for virtual humans that can interact with lectsacommunicationactandthenchoosesemotionalexpression people in virtual worlds, we have implemented an Army peace- based on some desired impact it will have on the user. For ex- keepingscenario,whichhasbeenviewedbyseveralhundredpeo- ample,Cosmo[10]makesaone-to-onemappingbetweenspeech pleandwasfavorablyviewedbymanydomainexperts[21]. As actsandemotionalbehaviorsthatreinforcetheact(congratulatory the simulation begins, a human user, playing the role of a U.S. actstriggeranadmirationemotiveintentthatisconveyedwithap- Army lieutenant, finds himself in the passenger seat of a simu- plause). BallandBreese[2]intendtoconveyasenseofempathy latedvehiclespeedingtowardsaBosnianvillagetohelpaplatoon byexpressingbehaviorsthatmirrortheassessedemotionalstateof introuble.Suddenly,heroundsacornertofindthatoneofhispla- theuser. PoggiandPelachaud[16]usefacialexpressionstocon- toon’svehicleshascrashedintoacivilianvehicle,injuringalocal veytheperformativeofacommunicationact,showing“potential boy(Figure1).Theboy’smotherandanArmymedicarehunched anger”tocommunicatethattheagentwillbeangryifarequestis overhim,andasergeantapproachesthelieutenanttobriefhimon notfulfilled. the situation. Urgent radio calls from theplatoon downtown, as wellasoccasionalexplosionsandweaponsfirefromthatdirection, In contrast, appraisal theories focus on the apparent evaluative suggestthatthelieutenantsendhistroopstohelpthem.Emotional function emotions play in human reasoning. Appraisal theories pleasfromtheboy’smother,aswellasagrimassessmentbythe view emotion asarising fromsome assessment of an agent’s in- medicthattheboyneedsamedevacimmediately,suggestthatthe ternal state vis-a`-vis its external environment (e.g., is this event lieutenantinsteadusehistroopstosecurealandingzoneforthe contrary to my desire?). Such appraisals can be used to guide medevac helicopter. The lieutenant carries on a dialog with the decision-making andbehaviorselection. Forexample, Beaudoin sergeantandmedictoassessthesituation,issueorders(whichare [3]usesthemtosetgoalprioritiesandguidemeta-planning.They carriedout bythesergeant through foursquads ofsoldiers), and canalsoserveasabasisforcommunicatinginformationaboutthe askforsuggestions. Hisdecisionsinfluencethewaythesituation agent’sassessment, though not intheintentional wayviewedby unfolds,culminatinginaglowingnewsstorypraisinghisactions communicative models. Appraisalmethodsareuseful forgiving or ascathing newsstory exposing the flawsinhisdecisions and coherence to an agent’s emotional dynamics that can be lacking describingtheirsadconsequences. inpurelycommunicativemodels. Thiscoherenceisessentialfor conveyingasenseofbelievabilityandrealism[13]. Whilethecurrentimplementationofthescenarioofferslittleflex- ibilitytothelieutenant,itprovidesarichtestbedforourresearch. Theseapproachesarecomplementary,thoughfewapproacheshave Currently, the mother, medic and sergeant are implemented as considered how to combine them into a coherent approach. An Steveagents.Allothervirtualhumans(acrowdoflocalsandfour exceptionisIPD[11],whichrepresentsbothappraisedemotional squadsofsoldiers)arescriptedcharacters. Wewillillustrateele- state and communicative intent and selectively expresses one or mentsofourdesignusingthemother,sinceheremotionsareboth the other based on a simple threshold model. In the discussion themost dramatic inthescenario andthe most crucial for influ- thatfollowswewillfocusontheproblemofemotionalappraisal. encingthelieutenant’sdecisionsandemotional state. Whileour currentimplementationofthemotherincludesapreliminaryinte- 4. MODELINGEMOTIONS grationof Steve, E´mile, and IPD,thefocusof thispaper isona moregeneralintegrationthatgoesbeyondthatimplementation. 4.1 Emotional State Ourprimarydesigngoalistocreategeneralmechanismsthatare 3. RELATED WORK nottiedtothedetailsofanyspecificdomain.Thus,emotionalap- Severalresearchershaveconsideredcomputationalmodelsofemo- praisals do not refer to features of the environment directly, but tiontodrivethebehaviorofsyntheticcharacters.Theseapproaches referenceabstractfeaturesoftheagent’sinformationprocessing. canberoughlycharacterizedasbeingeithercommunicationdriven Appraisalsandre-appraisalsaretriggeredreflexivelybychanges in an agent’s mental representations, which are in turn changed allassessmentoftheagent’semotionalstateandareusedtodrive bygeneralreasoningmechanisms. Externaleventscanindirectly the physical focus modes discussed below. Thismodel does yet impacttheseinternalrepresentations.Forexample,theagentmay supportsomeofthesubtledynamicsdisplayedbyIPDwhichcan perceive an event (e.g., soldiers are leaving the accident scene), convey an overall tone yet make agile transitions between con- andthenageneralreasoningmechanism(e.g.Steve’staskreason- veyedemotionalstate. Weareexploringsomedifferentstrategies ing) infers the consequences of this event (e.g., the departure of forachievingthiseffect,includingaggregatingsubsetsofappraisal troopsviolatesthepreconditionofthemtreatingmychild).Alter- framesondemanddependingofthecurrentdiscoursestate,orby natively, these representations may change astheresult of inter- addinganotheraggregatelayerthatincorporatesadifferentdecay naldecision-making(e.g.,forminganintentiontoviolateasocial rateintoappraisalintensities. obligation). Appraisal frameskey off of features of these repre- sentations(e.g.,thenegativeimpactofapreconditionviolationon goalprobabilityortheviolationofthesocialnormofmeetingyour 4.2 The Effect ofEmotions onBehavior obligations). Thedynamicsofbehaviorandemotionalexpression Theagentswedesignincorporateawiderangeofoutwardbehav- istherebydrivenbythemechanismsthatupdateanagent’smental iors in order to interact believably with the environment as well representations. as other agents and humans. Their bodies have fully articulated limbs,facialexpressions, andsensoryapparatus. Theycanmove TheE´milesystem[8]illustratesthisapproachbyusingadomain- intheenvironment,manipulateobjectsanddirecttheirgazeinap- independentplanningapproachtorepresentmentalstateandmake propriateways. Theyarecapableofrich,multi-modalcommuni- inferences, but E´mileis therefore largely restricted to appraisals cationthatincorporatesbothverbalbehaviorsaswellasnonverbal relatedtoanagent’sgoalsandchangestotheprobabilityofgoal behaviors.Inaddition,theyhavefacialexpressions,bodypostures attainment. IPD [11] explores amuchricher notionof appraisal andtheabilitytoperformvariouskindsofgestures.Thekeychal- includingavarietyofemotionalcopingstrategiesandaricherno- lengefortheagentdesignistomanagethisflexibilityintheagent’s tionofsocialrelationshipsandegoidentity,butisimplementedin physicalpresenceinawaythatconveysconsistentemotionalstate termsof alessgeneral inference mechanism. Inour joint work, andindividualdifferences.Toaddressthesechallenges,werelyon webegintobridgethegapbetweenbybuttressingE´mile-styleap- awiderangeofworkinhumanemotion,socialbehavior,clinical praisalswithabroaderrepertoireofinferencestrategies. Initially, psychology,animationandtheperformancearts. weareaugmentingplan-basedappraisalswiththoserelatedtodi- alog,incorporatingmethodsforrepresentingandupdatingthedis- Forexample,psychologicalresearchonemotionrevealsitsperva- coursestateandsocialobligationsthatarisefromdiscourse[15]. siveimpactonphysicalbehaviorsuchasfacialexpressions,gaze Thesestructurescanthenserveasthebasisforamoreuniformrep- andgestures[1,5,6].Thesebehaviorssuggestconsiderableinfor- resentation of social commitments and obligations beyond those mation about emotional arousal, attitudes and attention. For ex- strictlyrelatedtodialog,andthatinteractwithplan-basedreason- ample,depressedindividualsmayaverttheirgazeanddirecttheir ing(e.g. oneisobligatedtoredresswrongs). Thiswillallowthe gesturesinwardtowardstheirbodiesinself-touchingbehaviors. systemtomodelsomeoftherichdynamicsofemotionsrelating todiscoursethatIPDsupports,butdosoinamorestructuredway. Notethatsuchmovementsalsoservetomediatetheinformation available to the individual. For example, if a depressed individ- E´milecontains asetof recognitionrulesthat scananagent’sin- ual’sheadislowered,thisalsoregulatestheinformationavailable ternal representations and generate an appraisal frame whenever tothe individual. Orientingon anobject of fear or anger brings certainfeaturesarerecognized. Forexample,whenevertheagent theobjecttothefocusofperceptualmechanisms,whichmayhave adoptsanewgoal(orisinformedofagoalofsomeotheragent), indirectinfluencesoncognitionandcognitiveappraisalbyinflu- framesarecreatedtotrackthestatusofthatgoal. Eachappraisal encingthecontentofworkingmemory. Evenasoothingbehavior frame describes the appraised situation in terms of a number of likerubbinganarmmayservetomanagewhatapersonattendsto specificfeaturesproposedbytheOCCmodel[14]. Theseinclude [7]. the point of view from which the appraisal is formed, thedesir- abilityofthesituation,whetherthesituationhascometopassoris These findings provide a wealth of data to inform agent design onlyapossibility,whetherthesituationmeritspraiseorblame(we butoftenleaveopenmanydetailsastohowalternativebehaviors donot model theappealingness ofdomain objects, an important aremediated. Theagenttechnology weuseallowsonetocreate factorinOCC).Forexample,ifsomeonethreatensmygoal,from richphysical bodiesforintelligentcharacterswithmanydegrees myperspectivethisisanundesirablepossibilitythatmeritsblame. of physical movement. This forces one to directly confront the Basedonthesettingofthevariousfeatures,oneormore“emotion problemofemotionalconsistency. Forexample,an“emotionally instances” are generated according to OCC. E´mile incorporates depressed” agent might avert gaze, be inattentive, perhaps hug- a decision-theoretic method for assigning intensities to these in- gingthemselves. However,ifinsubsequentdialogtheagentused stances. Intensityvalues decay over time but are “re-energized” strongcommunicativegesturessuchasbeats[12],thenthebehav- whenever the underlying representational structures are manipu- ior might not “read” correctly. Thisissue is well known among lated. animators. Forexample, FrankThomasand OllieJohnston, two of Disney’s “grand old men,” observe that “the expression must Individualemotioninstancesareaggregatedinto“buckets”corre- becapturedthroughoutthewholebodyaswellasintheface”and spondingtoemotionsofthesametype. Thus,threatstomultiple “any expression will beweakened greatly if it islimitedonly to goalswillbeaggregatedintoanoverallleveloffear.Inthissense, theface,anditcanbecompletelynullifiedifthebodyorshoulder E´milecontainsatwo-levelrepresentationofemotionalstate.Emo- attitudeisinanywaycontradictory”[22]. tioninstancesarethelevelatwhichsomesemanticconnectionis still retained between the emotion and the mental structures un- Implicitinthesevariousconcernsisthattheagenthaswhatamounts derlying the appraisal. Thus, the agent has the capability of an- to a resource allocation problem. The agent has limited physi- sweringquestionsaboutitsemotionstate(e.g.,“Whyareyouso cal assets, e.g., two hands, one body, etc. At any point in time, angry?”). Theaggregatebucketsroughlycorrespondtotheover- theagent mustallocatetheseassetsaccordingtoavarietyofde- mands,suchasperformingatask,communicating,oremotionally and physical dynamics. The agent is also more attentive to and soothingthemselves. Forinstance,theagent’sdialogmaybesug- responsivetoeventsintheenvironment. gestiveofaspecificgesturefortheagent’sarmsandhandswhile theemotionalstateissuggestiveofanother.Theagentmustmedi- Transition between modes is currently based on emotional state atebetweenthesealternativedemandsinafashionconsistentwith derived from the appraisal model. High levels of sadness, de- theirgoalsandtheiremotionalstate.Fromthehumanparticipant’s creasedhope,orguilt,inabsolutetermsandrelativetootheremo- perspective,weexpectthisconsistencyintheagent’sbehaviorto tionalbuckets,inducestransitionsawayfromCommunicativeFo- supportbelievabilityandinterpretability. cus and towards Body Focus. Increased hope or anger induces transitionstowardsCommunicativeFocus.Thetransitionsarede- ToaddressthisproblemwerelyonthePhysicalFocusmodel[11], signed to be“sticky” sothat the agent does not readilypop into acomputationaltechniqueinspiredbyworkonnonverbalbehavior andthenoutofamode. TransitionalFocus,truetoitsname,lies inclinical settings[7] and Lazarus’s [9] delineation of emotion- betweentheothertwomodes. However,extremeemotionalstate directedversusproblem-directedcopingstrategies. ThePhysical canbypassit. Focusmodelbasesanagent’sphysicalbehaviorintermsofwhat the character attends to and how they relate to themselves and Groupingbehaviorsintomodesattemptsto(a)mediatecompeting theworldaroundthem,specificallywhethertheyarefocusingon demandsonanagent’sphysicalresourcesand(b)coalescebehav- themselves and thereby withdrawing from the world or whether iorsintogroupsthatprovideaconsistentinterpretationabouthow theyarefocusing onthe world, engaging it. Inother words, the theagentisrelatingtoitsenvironment. Itisdesignedtodothisin model bases physical behavior on how the agent is choosing to afashionconsistentwithemotionalstate,withtheintentthatitbe copewiththeworld. generalacrossagents.However,realismalsorequiresthatspecific behaviorswithineachmodeincorporateindividualdifferences,as The model organizes possible behaviors around a set of modes. inhumanbehavior. Forexample,wewouldnotexpectamother’s Behaviorscanbeinitiatedviarequestsfromtheplanner/executor repertoireofgesturestobeidenticaltothatofanarmysergeant. orstartedspontaneouslywhenthebodyisnototherwiseengaged. Indeed, the current implementation is a step en route to a fuller Atanypointintime,theagentwillbeinauniquemodebasedon renderingofacopingmodel:amodelofhowtheagentchoosesto thecurrentemotionalstate. copewithemotionalstressthatincludesbothemotionalappraisal aswellasindividualdifferences(personality). In the current work, we model three modes of physical focus: BodyFocus,TransitionalFocusandCommunicativeFocus(asop- 5. CONCLUSION posedtothefivemodesdiscussedin[11]andidentifiedinFreed- The challenge of this research is not an easy one. We seek to man’swork).Thefocusmodeimpactsbehavioralongseveralkey createvirtualhumanswhosebehaviorisabelievablerenditionof dimensions,includingwhichbehaviorsareavailable,howbehav- human-likebehavior,isresponsivetodynamicallyunfoldingsce- iorsareperformed, how attentivetoexternal events theagent is, nariosandisinterpretableusingthesamecuesthatpeopleuseto andwhichactionstheagentselectsandhowitspeaks. understandeachother. Akeyfactorforsuccesswillbeafaithful renderingofthecauseofemotionanditsoutwardmanifestation. TheBodyFocusmodemodelsaself-focusedattention,awayfrom Toaddressthatchallenge,wehavebeguntointegratecomponents theworld,andisalsodesignedtorevealconsiderabledepression fromthreeprevioussystems: Steve’svirtualhumanarchitecture, or guilt. Theprevalent typeof gestures availabletotheagent in E´mile’semotional appraisal and IPD’scomplementary model of this mode are self-touching gestures, or adaptors [5], which in- emotionalappraisalanditsimpactonphysicalbehavior. volvehand-to-body gesturesthatappear soothing(e.g., rhythmic strokingofforearm)orself-punitive(e.g.,squeezingorscratching Aninitialintegrationhasalreadybeenrealizedandimplemented of forearm). Conversely, communicative gestures such as deic- inthemothercharacterofourArmypeacekeepingscenario.How- ticorbeatgestures[12]aremutedinthismodebothintermsof ever, our integrationefforts areongoing. Our current effortsare their numbers and dynamic extent. Indeed, overall gesturing in ongeneralizingthatintegrationandincorporatingmoreoftheIPD thismodeismutedinitsdynamics. Also,theagentislessatten- social-based appraisals into the general inferencing mechanisms tivetotheenvironmentandinparticularexhibitsconsiderablegaze inSteveandE´mile. aversionanddownwardlookinggaze.Theagent’sverbalbehavior isinhibited and marked by pauses. Intermsof action selection, 6. ACKNOWLEDGMENTS theagenthasareducedpreferenceforcommunicativeacts. This research was funded by the Army Research Institute under TransitionalFocusindicatesalessdividedattention,lessdepres- contractTAPC-ARI-BR,theNationalCancerInstituteundergrant sionorguilt,aburgeoningwillingnesstotakepartintheconversa- R25CA65520-04,andtheU.S.ArmyResearchOfficeundercon- tion,milderconflictswiththeproblemsolvingandacloserrelation tractDAAD19-99-C-0046. tothelistener. Thismode’sgesturingismarkedbyhand-to-hand gestures (such as rubbing hands or hand fidgetiness) and hand- 7. REFERENCES to-object gestures, such as playing with a pen. There are more [1] M.ArgyleandM.Cook.GazeandMutualGaze. communicative gestures in this mode but they are still muted or CambridgeUniversityPress,Cambridge,1976. stilted. [2] G.BallandJ.Breese.Emotionandpersonalityina Finally,CommunicativeFocusisdesignedtoindicateafullwill- conversationalagent.InJ.Cassell,J.Sullivan,S.Prevost, ingness toengage the world in dialog and action. In thismode, andE.Churchill,editors,EmbodiedConversationalAgents. theagent’sfullrangeofcommunicativegesturesareavailable.On MITPress,Cambridge,MA,2000. theotherhand,adaptorsaremutedandoccuronlywhentheagent is listening or otherwise not occupied. In contrast, the commu- [3] L.Beaudoin.GoalProcessinginAutonomousAgents.PhD nicativegesturesaremoreexaggerated,bothintermsofnumbers thesis,UniversityofBirmingham,1995.CSRP-95-2. [4] L.Berkowitz.CausesandConsequencesofFeelings. [20] J.RickelandW.L.Johnson.Task-orientedcollaboration CambridgeUniversityPress,2000. withembodiedagentsinvirtualworlds.InJ.Cassell, J.Sullivan,S.Prevost,andE.Churchill,editors,Embodied [5] P.EkmanandW.V.Friesen.Therepertoireofnonverbal ConversationalAgents.MITPress,Cambridge,MA,2000. behavior:Categories,origins,usage,andcoding.Semiotica, 1:49–97,1969. [21] W.Swartout,R.Hill,J.Gratch,W.Johnson,C.Kyriakakis, C.LaBore,R.Lindheim,S.Marsella,D.Miraglia, [6] P.EkmanandW.V.Friesen.Constantsacrossculturesinthe B.Moore,J.Morie,J.Rickel,M.Thie´baux,L.Tuch, faceandemotion.PersonalityandSocialPsychology,17(2), R.Whitney,andJ.Douglas.Towardtheholodeck: 1971. Integratinggraphics,sound,characterandstory.In [7] N.Freedman.Theanalysisofmovementbehaviorduring ProceedingsoftheFifthInternationalConferenceon theclinicalinterview.InA.SiegmanandB.Pope,editors, AutonomousAgents,2001. StudiesinDyadicCommunication,pages177–210. [22] F.ThomasandO.Johnston.TheIllusionofLife:Disney PergamonPress,NewYork,1972. Animation.WaltDisneyProductions,NewYork,1981. [8] J.Gratch.E´mile:Marshallingpassionsintrainingand education.InProceedingsoftheFourthInternational ConferenceonAutonomousAgents,pages325–332,New York,2000.ACMPress. [9] R.Lazarus.EmotionandAdaptation.OxfordPress,1991. [10] J.C.Lester,J.L.Voerman,S.G.Towns,andC.B. Callaway.Deicticbelievability:Coordinatinggesture, locomotion,andspeechinlifelikepedagogicalagents. AppliedArtificialIntelligence,13:383–414,1999. [11] S.C.Marsella,W.L.Johnson,andC.LaBore.Interactive pedagogicaldrama.InProceedingsoftheFourth InternationalConferenceonAutonomousAgents,pages 301–308,NewYork,2000.ACMPress. [12] D.McNeill.HandandMind:WhatGesturesRevealabout Thought.UniversityofChicagoPress,1992. [13] W.S.NealReilly.BelievableSocialandEmotionalAgents. PhDthesis,SchoolofComputerScience,CarnegieMellon University,Pittsburgh,PA,1996.TechnicalReport CMU-CS-96-138. [14] A.Ortony,G.Clore,andA.Collins.TheCognitive StructureofEmotions.CambridgeUniversityPress,1988. [15] M.PoesioandD.Traum.Towardsanaxiomatizationof dialogueacts.InJ.HulstijnandA.Nijhol,editors, ProceedingsoftheTwenteWorkshopontheFormal SemanticsandPragmaticsofDialogues(13thTwente WorkshoponLanguageTechnology),pages207–222, Enschede,1998. [16] I.PoggiandC.Pelachaud.Emotionalmeaningand expressioninperformativefaces.InInternationalWorkshop onAffectinInteractions: TowardsaNewGenerationof Interfaces,Siena,Italy,1999. [17] J.Rickel,J.Gratch,R.Hill,S.Marsella,andW.Swartout. StevegoestoBosnia:Towardsanewgenerationofvirtual humansforinteractiveexperiences.InAAAISpring SymposiumonArtificialIntelligenceandInteractive Entertainment,March2001. [18] J.RickelandW.L.Johnson.Animatedagentsfor proceduraltraininginvirtualreality:Perception,cognition, andmotorcontrol.AppliedArtificialIntelligence, 13:343–382,1999. [19] J.RickelandW.L.Johnson.Virtualhumansforteam traininginvirtualreality.InProceedingsoftheNinth InternationalConferenceonArtificialIntelligencein Education,pages578–585.IOSPress,1999.