Arguing to Motivate Decisions TheresearchreportedhereispartoftheInteractiveCollaborativeInformationSystems (ICIS)project(http://www.icis.decis.nl/),supportedbytheDutchMinistryof EconomicAffairs,grantnr: BSIK03024. SIKSDissertationSeriesNo. 2011-33 TheresearchreportedinthisthesishasbeencarriedoutundertheauspicesofSIKS,the DutchResearchSchoolforInformationandKnowledgeSystems. c 2011T.L.vanderWeide (cid:13) PrintedbyWo¨hrmannPrintService,Zutphen LATEX templatebySusanvandenBraak ISBN978-90-393-56494 Arguing to Motivate Decisions Argumentatie voor het Motiveren van Beslissingen (meteensamenvattinginhetNederlands) PROEFSCHRIFT terverkrijgingvandegraadvandoctoraandeUniversiteitUtrechtopgezag vanderectormagnificus,prof.dr.G.J.vanderZwaan,ingevolgehetbesluit vanhetcollegevoorpromotiesinhetopenbaarteverdedigenopmaandag10 oktober2011desmiddagste12.45uur door Thomas Leonardus van der Weide geborenop27juli1982teNijmegen Promotoren: Prof. dr. J.-J.Ch. Meyer Prof. dr. mr. H.Prakken Co-promotoren: Dr. F.P.M.Dignum Dr. G.A.W.Vreeswijk 1 Contents 1 Introduction 1 1.1 IdentifyingtheBestDecision . . . . . . . . . . . . . . . . . . . . . . . . . . 2 1.1.1 DecisionTheory . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 1.1.2 Value . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 1.1.3 ResearchQuestionsonArgumentation. . . . . . . . . . . . . . . . . 5 1.2 ResearchQuestionsonDialogues . . . . . . . . . . . . . . . . . . . . . . . 7 1.3 OverviewoftheThesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 2 Background 11 2.1 Argumentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 2.1.1 ArgumentationSystems . . . . . . . . . . . . . . . . . . . . . . . . 12 2.1.2 AttackbetweenArguments . . . . . . . . . . . . . . . . . . . . . . . 17 2.1.3 ArgumentStrengthandDefeat . . . . . . . . . . . . . . . . . . . . . 18 2.1.4 ArgumentationFrameworks . . . . . . . . . . . . . . . . . . . . . . 20 2.1.5 RationalityofConclusions . . . . . . . . . . . . . . . . . . . . . . . 22 2.2 Accrual . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 2.3 Meta-LevelArgumentation . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 2.3.1 Meta-ArgumentationSystem . . . . . . . . . . . . . . . . . . . . . . 29 2.3.2 Meta-ArgumentationTheories . . . . . . . . . . . . . . . . . . . . . 31 2.3.3 Meta-ArgumentationFramework . . . . . . . . . . . . . . . . . . . 33 2.3.4 SummaryonArgumentation . . . . . . . . . . . . . . . . . . . . . . 37 2.4 DecisionTheory. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 2.4.1 Value . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 2.4.2 Multi-AttributeUtilityTheory . . . . . . . . . . . . . . . . . . . . . 39 2.4.3 DecisionAnalysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 2.5 AbstractValues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 2.5.1 WhatAreAbstractValues? . . . . . . . . . . . . . . . . . . . . . . . 42 2.5.2 HumanAbstractValues. . . . . . . . . . . . . . . . . . . . . . . . . 43 2.5.3 PrioritiesbetweenValues . . . . . . . . . . . . . . . . . . . . . . . . 43 2.5.4 AbstractValuesandArgumentation . . . . . . . . . . . . . . . . . . 45 i CONTENTS 3 ConceptualFrameworkForValue 49 3.1 AlternativesandAssignments . . . . . . . . . . . . . . . . . . . . . . . . . 50 3.2 Dyadic/ComparativeValue . . . . . . . . . . . . . . . . . . . . . . . . . . 51 3.2.1 DesiredProperties . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 3.2.2 ValueinPracticalReasoning . . . . . . . . . . . . . . . . . . . . . . 54 3.2.3 GeneralizingValueStatements . . . . . . . . . . . . . . . . . . . . . 55 3.3 ValueTrees . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60 3.3.1 InfluenceBetweenPerspectives . . . . . . . . . . . . . . . . . . . . 61 3.3.2 TransitivityofInfluence . . . . . . . . . . . . . . . . . . . . . . . . 65 3.3.3 Importance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66 3.4 Monadic/ClassificatoryValue . . . . . . . . . . . . . . . . . . . . . . . . . 68 3.4.1 PropertiesofMonadicValuePredicates . . . . . . . . . . . . . . . . 69 3.4.2 JustifyingMonadicValue . . . . . . . . . . . . . . . . . . . . . . . 71 3.4.3 InPracticalReasoning . . . . . . . . . . . . . . . . . . . . . . . . . 72 3.5 ChapterSummary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73 4 Perspective-BasedValueModel 75 4.1 Assignments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75 4.2 Perspectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 4.2.1 InfluenceBetweenPerspectives . . . . . . . . . . . . . . . . . . . . 80 4.2.2 RelativeImportance . . . . . . . . . . . . . . . . . . . . . . . . . . 81 4.3 MonadicEvaluations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82 4.4 RunningExample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85 4.5 ChapterSummary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89 5 ArgumentationaboutPerspective-BasedValue 91 5.1 Object-LevelArgumentation . . . . . . . . . . . . . . . . . . . . . . . . . . 92 5.1.1 ArgumentationSystemforPerspective-BasedValue . . . . . . . . . 92 5.1.2 ArgumentationTheories . . . . . . . . . . . . . . . . . . . . . . . . 96 5.2 Meta-LevelArgumentation . . . . . . . . . . . . . . . . . . . . . . . . . . . 100 5.2.1 Meta-ArgumentationSystemforPV . . . . . . . . . . . . . . . . . . 100 5.2.2 Meta-ArgumentationTheories . . . . . . . . . . . . . . . . . . . . . 102 5.3 RunningExample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103 5.4 ChapterSummary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106 6 PracticalReasoning 109 6.1 Deliberation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110 6.1.1 DecisionContext . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111 6.1.2 JustificationofAchievementGoals . . . . . . . . . . . . . . . . . . 111 6.1.3 JustificationofAvoidanceGoals . . . . . . . . . . . . . . . . . . . . 113 6.2 Means-EndReasoning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114 6.2.1 OutcomesofAlternatives. . . . . . . . . . . . . . . . . . . . . . . . 114 6.2.2 JustificationofDecisions . . . . . . . . . . . . . . . . . . . . . . . . 115 6.2.3 StrengthofArgumentsSchemes . . . . . . . . . . . . . . . . . . . . 117 ii CONTENTS 6.3 Formalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118 6.3.1 Perspective-BasedArgumentationforPracticalReasoning . . . . . . 118 6.3.2 Meta-Argumentation . . . . . . . . . . . . . . . . . . . . . . . . . . 123 6.4 RunningExample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125 6.5 RelatedWork . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130 6.5.1 Atkinsonetal. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130 6.5.2 DecisionRules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132 6.5.3 AmgoudandPrade . . . . . . . . . . . . . . . . . . . . . . . . . . . 133 6.6 ChapterSummary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135 7 ADialogueFrameworkforSupportingDecisions 137 7.1 RelatedWork . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138 7.1.1 DialogueTypes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138 7.1.2 DeliberationDialogues . . . . . . . . . . . . . . . . . . . . . . . . . 139 7.1.3 RelevanceinDialogues . . . . . . . . . . . . . . . . . . . . . . . . . 140 7.2 DialogueFramework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141 7.2.1 CommunicationLanguage,DialogueMoves,andDialogues . . . . . 141 7.2.2 DecisionSupportDialogues . . . . . . . . . . . . . . . . . . . . . . 145 7.2.3 AProtocolforDecisionSupport . . . . . . . . . . . . . . . . . . . . 146 7.3 RevisingBeliefs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155 7.4 RunningExample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157 7.5 ChapterSummary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161 8 MoveSelectionwithMultipleCriteria 163 8.1 CriteriaforMoveSelection . . . . . . . . . . . . . . . . . . . . . . . . . . . 164 8.1.1 Value-basedCriteria . . . . . . . . . . . . . . . . . . . . . . . . . . 164 8.1.2 AcceptabilityofArguments . . . . . . . . . . . . . . . . . . . . . . 165 8.1.3 Interest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165 8.2 RunningExample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166 8.3 ChapterSummary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174 9 Conclusion 177 9.1 IdentifyingTheBestDecision . . . . . . . . . . . . . . . . . . . . . . . . . 178 9.2 SupportingaDecision. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181 9.3 Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183 9.4 RecommendationsforFutureWork. . . . . . . . . . . . . . . . . . . . . . . 184 References 187 Samenvatting 195 Dankwoord 197 SIKSDissertationSeries 199 iii 1 Introduction Consider a critical and complex decision situation such as being a fire commander in a sit- uationwherethereisafireinafactorycontainingtoxicchemicalsandwithpeopletrapped inside.Thefirecommanderhastodecidewhatcourseofactionhastobetaken.Thisdecision impactsnotonlythelivesofthevictimstrappedinsidethefactory,butalsothelivesofthefire fightersthatmayhavetorescuethem. Moreover,itispossiblethatthetoxicchemicalsinside the factory will escape into the environment. Because of the complexity of such decisions andthelimitedamountoftimethatisavailable,itisbothusefulandinterestingtoinvestigate howsuchdecisionscanbesupported. Thepurposeofadecisionsupportsystemistoaidadecision-makerindeterminingwhat decisionhe1shouldtake.Decision-makingroughlyrequiresdetermining(1)whatalternatives are available; (2) in what outcome each alternative results; and (3) what expected outcome ispreferredthemost. Decisionscanbesupportedinvariousways,butinthisthesiswewill focus on how to support determining which expected outcome is preferred the most. It is natural for people to use argumentation to reason about and to communicate what decision is the best (Shafir et al., 1997). To support decision making in a natural way, we therefore addressthefollowingmainresearchquestioninthethesis. MainResearchQuestion: Howcanargumentationbeusedtosupportdecisionmakingincomplex scenarios? RunningExample In this thesis we will use a ‘running example’ to demonstrate what is needed in existing workandhowourproposedframeworkimprovesuponexistingwork. Therunningexample is in the context of so-called ‘serious games’. Although computer games are mainly used to entertain, serious games are increasingly used for training and education (Michael and Chen, 2005). For example, Johnson et al. (2005) use serious games to teach language and Schurr et al. (2005) use them as a training tool for incident commanders. In particular, we are interested in serious games where students train to become fire commanders and learn howtomakedecisionsincomplexscenarios. Whenthereisafire,afirecommanderhasto 1Forbrevity,weuse‘he’toreferto‘heorshe’and‘his’torefer‘hisorher’. 1 INTRODUCTION 1.1 consider all relevant information and make a decision about what the fire fighters will do. In the computer game, the student plays the character of fire commander in a virtual world inwhichtherepotentiallyaremultiplefires. Thestudenthastodecidewhatthefirefighters shoulddo. Therearemanytypesoffiresthatafirecommandermayhavetodealwith. Some firesarerelativelysimple(e.g.,aburningcarinaremotearea),butothersarecomplicated. Example1.1(RunningExample: Fireinafactory) There is a fire in a factory where toxic chemicals are stored. Inside the factory, there are several people whose lives are in danger and who need help to get out. Because the fire is damaging the factory, there is an increasedriskthatthetoxicchemicalsescapeanddamagetheenvironmentandhealthofthe people in the surrounding area. The fire commander arrives at the scene and has to decide whatcourseofactionshouldbetaken. Althoughsuchcomplexsituationsdonotoccurfrequently,thepossibleconsequencesare serious. It is thus important that fire commanders train how to act in such scenarios. The complexityofsuchdecisionsarisesroughlyfromthefollowingproblems. Ontheonehand, there is uncertainty w.r.t. the outcome of actions and what the current situation exactly is (Bharosa et al., 2010). On the other hand, there are many different aspects to what makes onedecisionbetterthananotherandalltheseaspectsaredifferentbynature. Consequently, determining what decision is better is difficult, even if there is certainty concerning their outcomes. Forexample,intherunningexamplethefollowingperspectivesmaybeimportant to a decision maker: the safety of the victims trapped inside, the safety of the people in thesurroundingarea,andthesafetyofthefirefightingpersonnel,butalsotheimpactonthe environment, the costs and trouble for people e.g., caused by when a road is closed. When determining what decision to take, it is possible that one or multiple of these aspects are forgotten or that a mistake is made. To help a decision maker understand why a certain decision is better than another decision, a decision support system might argue to persuade the decision maker why the system believes that a certain decision should be taken. In this thesis,wefocusontraininghowtodeterminewhatdecisionisbetter. Supporting a user to make the best decision involves two aspects: the system needs to knowwhatdecisionisthebestandthesystemneedstobeabletosupporttheusereffectively. Section1.1zoomsinonhowthesystemcandeterminewhatdecisionisthebestfortheuser andSection1.2furtherinvestigateshowausercanbesupported. Theintroductionisended by summarizing what we have discussed in the introduction and giving an overview of the restofthethesis. 1.1 Identifying the Best Decision Tosupportadecisionmakerinmakingthebestdecision,adecisionsupportsystemneedsto havesomeideaaboutwhatdecisionisbest. Wewillnowfurtherinvestigatewhatisrequired forthisrequirement. 1.1.1 DecisionTheory Onepossibletechniquethatcouldbeusedtoreasonaboutandexplainwhyacertaindecision should be taken is utility theory. The interdisciplinary area of decision theory studies how 2