A Collaborative Development Environment for Design–Oriented Critiquing Systems

ACollaborativeDevelopmentEnvironmentforDesign–OrientedCritiquingSystems

MaritaDuecker

BerndGutkaufStefanieThiesC–LABAVL–ListGmbHHeinzNixdorfInstitutFuerstenallee11

HansListPlatz1Fuerstenallee11

33102Paderborn,Germany

8020Graz,Austria33102Paderborn,Germanymary@c–lab.de

bernd.gutkauf@avl.comthies@hni.uni-paderborn.de

Abstract

Design–orientedcritiquingsystemshavesuccessfullydemonstratedtheircapabilitytoaidusersindesigningar-tifacts.Theycanbeseenaspersonalassistantshelpingtoreflectonaparticulardesignandtoimproveit.Fordo-ingso,design–orientedcritiquingsystemsmakeknowledgeandexpertiseofdomainexpertsavailabletoendusersofauthoringtools.Unfortunately,itisaquitedemandingandtimeconsumingtasktocompiletheexpertiseofdomainex-pertsandtomaintainit.

Wepresentacollaborativedevelopmentenvironmentforsupportingdomainexpertsinbuildingdesign–orientedcri-tiquingsystems.Inthisarticlewefocusonassistingdo-mainexpertsinbuildingaconsistentandpreferablycom-pleteknowledgebaseofadomain.Aknowledgebaseformsthebasisofacritiquingsystem.Well–structurednegotia-tionprocessesclearuppredictabledisagreementsarisingwhendomainexpertsfromvariousdisciplinescompiletheirknowledgecollaboratively.Consideringthatexpertsarenotnecessarilyfamiliarwithprogrammingweprovidemeansforrapidprototypingandmaintenanceofdesign–orientedcritiquingsystems.

1.Introduction

Withtheincreasingperformanceofcomputers,author-ingtools,e.g.,forcreatingmultimediaandWWWpresen-tations,andscientificvisualizationsaregettingmoreandmorecomplex.Design–orientedcritiquingsystemshavebeendevelopedforsupporting(end)usersofauthoringtoolsintheirdesigntasksandthereforealsoformanagingtheap-plication’scomplexity.1Critiquingsystemsassistusersin

0-7695-0493-0/00 $10.00 (c) 2000 IEEE1

Proceedings of the 33rd Hawaii International Conference on System Sciences - 2000

Figure1.UserandExpertLevelofCritiquing.

tively.Jointviewsonknowledgeandjointmanipulationofknowledgeenableexpertstoimproveandensureaknowl-edgebase’squality.

Thispaperisorganizedasfollows.Thenextsectiondis-cussesrelatedworkinthefieldofnegotiationsinceourpa-perisfocussedonthenegotiationprocessesbetweenex-perts.Section3introducescritiquingsystemsandtermsusedinthiscontext.Thereafter,weintroduceourmainob-jectiveswithregardtocompilingknowledgeinSection4followedbythepresentationofourconcepts.OurconceptsincludeageneralprocessmodelforcompilingknowledgeandstrategiesforsupportingnegotiationprocesseswhichisdescribedinSection5.Section6presentsarchitectureandcomponentsofoursystem.AfirstprototypeisdescribedinSection7whichgivesanoverviewofwhathasbeenreal-izedsofar.Thispaperendswithconclusionsandaspectsoffuturework.

2.RelatedWork

Acoupleofnegotiationsupportsystemshavebeenal-readyintroducedwhichcanberoughlyarrangedbytwoclasses.BuiandShakun[3]aswellasYuanetal.[19]de-scribenegotiationsupportsystemsaseithersolution–drivenorprocess–oriented.Asolution–drivennegotiationsupportsystemsupportsthenegotiatingpartiesincomingtoasolu-tionoranagreement.Possibleagreementsandalternativesolutionsareproposedbythesystem.Theseproposalsarederivedfromseveralmodelsonwhichnegotiationsmaybebasedupon.Examplesformodelsaregametheoreticmod-els,multiplecriteriamodels,negotiationanalysis,analyticmodelsorsimulationmodels.Aprocess–orientednegotia-tionsupportsystemdoesnotprovideanysolution.Itintendstosupportthenegotiationprocessbetweenthenegotiatingparties,startingwiththeplanningphaseandclosingwithan

achievedagreement.Incontrasttosolution–drivennegotia-tionsupportsystemsprocess–orientedonesmainlyaddresstwospheresofactivity:enrichedcommunicationchannelsbetweennegotiatingpartiesinordertoprovideinformationexchangeandcooperativework.Mostnegotiationsupportsystems,e.g.,HIPRE3+GroupLink[10],NEGOTIATOR[3],andINSS[14]aresolution–driven.CBSS[19]isanexampleforaprocess–orientednegotiationsupportsystem.HIPRE3+GroupLink[10]isagroupdecisionsupportsystem.Ifanegotiationsupportsystemssupportsnegotia-tionwithinagroup,thenitisaspecialkindofgroupdeci-sionsupportsystems[12].WithinanetworkedenvironmentHIPRE3+GroupLinkisabletosupportreal–timegroupmeetings.Eachusercansethis/herindividualpreferencesconcerningdifferentalternativesusingHIPRE3+.Thein-dividualpreferencesaredescribedthroughananalytichier-archyprocessmodelwhichisevaluatedbythesystem.Thisresultsinarankingofalternatives.Differentstrategiesal-lowtoinfluencetheindividualpreferences.Themainideainallproceduresisthatthegroupcanfocusthediscussionontheprioritizationswhichhavethehighestdifferencesinopinion.

NEGOTIATOR[3].ThenegotiationsupportsystemNEGOTIATORimplementstheevolutiveapproachtocon-flictresolutionwhichhasbeenintroducedbyBuiandShakun[3].Thisevolutiveapproachisbasedontwohierar-chyrelations.Onehierarchyrelationprovidesaframeworkwhichallowstospecifythegeneralproblemandtodetecttheparties’interests.Thesecondoneisaframeworkforfindingasolution.Problemrestructuringandfindingcom-promisesolutionsareincluded.Bothhierarchiesarebasedonmathematicalrelations.

INSS[14]isaweb–basednegotiationsupportsystemandsupportsthreebasicsteps.Apreparationphasefornegotiationincludestheanalysisofissuesandinterestsin-volved.Thepreferenceelicitationallowsuserstodeterminetheimportanceofeachissue.Duringtheconductphaseof-fersandmessagescanbecreated.Displaysprovideratingsbesideeachofferbasedonthepreferencesettings.Duringthepost–settlementperiodtherenegotiationofanagree-mentispossible.Preferencesettingsandtheagreementwhichhasbeenreachedareusedforanalyzingwhethertheagreementisanoptimalone.INSSalsomakessuggestionsforbettersolutions.

CBSS(CollectiveBargainingSupportSystem)[19]isaprocess–orientednegotiationsupportsystem.CBSSintendstosupporttwopossiblyspatiallydistributedpartiesinnego-tiatingthroughtheWeb.CBSSprovidessynchronousaids,suchascoordinationfacilitiesfornegotiationprocesses,ser-vicesformessageexchange,andtoolsforworkingoncom-mondocuments.AnegotiationitselfismainlystructuredaccordingtoGulliver’snegotiationprocessstages[8].Itdealswithfindingalocationforthenegotiationwhichisthe

0-7695-0493-0/00 $10.00 (c) 2000 IEEE2

Proceedings of the 33rd Hawaii International Conference on System Sciences - 2000

Webinthiscase.Apartfromthatthenegotiationprocessisroughlystructuredbythephasesforpreparingthenegotia-tion,settingtheagenda,discussingtheissuesandmakingafinalagreement.

AlthoughthisisonlyasampleofexistingnegotiationsupportsystemsitbecomesobviousthatmostsystemsarefocussedonfindingasolutionlikeHIPRE3+,NEGOTIA-TOR,andINSS.Incontrasttothesesystemsonlyafewareconcernedwithsupportingthenegotiationprocessit-self.JustasinCBSSourconceptsfornegotiationsupportareprocess–oriented.NegotiationsfollowthenegotiationprocessmodelbyKersten[13]whichisdescribedindetailinSubsection5.2.ItisderivedfromamodelproposedbyGulliveronwhichCBSSisbased.IncontrasttoCBSSweprefertostructureournegotiationprocessesaccordingtoKersten,becausehetakespresentandfutureuseofcom-putingandcommunicationtechnologiesintoconsideration.FurtherCBSSintendstosupportnegotiationsofgeneralas-pectscontributedbynegotiatingparties.Weaimatselec-tivelysupportingthediscussionofissuesarisingwhencom-pilingknowledgecollaboratively.Negotiationissueswillbepartlygivenbythesystemitself.

eachotherareanalyzedusingtheknowledgebasewhichisrepresentedbyrules.

Rules.Asimplifiedexampleshoulddemonstratehowrulesarebuilt.Rulesconsistofexternalfacts,conditionsandacritiquetext.Letusassumethataruleisspecifiedforcritiquingbusinessgraphs,moreprecisely,piechartscon-tainingneighboredpiesliceswhichhaveindistinguishablecolors.Inthiscasecolorsofpieslicesaretheexternalfacts.Therule’sconditionbecomestrueifthecolorsofneigh-boredpieslicesareindistinguishable.Thecritiquetextgivesexplanationswhycolorsshouldbedistinguishable.Ifneighboredpieslicesarebothcoloredblue,forexample,therule’sconditionbecomestrue.Thiscausesthecritiquetexttobeshowntotheuser.

ArtifactHierarchy.Rulesareevaluatedinacertainorderwhichisdependentonanartifacthierarchy.Alldesignob-jectsorartifactsofabusinessgraphlikebarchart,piechart,bar,drawingarea,coordinates,x–axis,color,etc.canbear-rangedinahierarchicalorderasitisshowninFigure3.

3.AnIntroductiontoCritiquingSystems

Wearegivingashortintroductiontocritiquingsystemsandtermsusedinthiscontext.GeneralinformationaboutcritiquingandcritiquingsystemsisgivenbyFischeretal.in[7]andbySilvermanin[17],forexample.Foridentifyingprerequisitesandoutliningthecontextofcritiquingsystemsweconsideracritiquingsystemforbusinessgraphdesignasanexample.AsketchofsuchasystemisshowninFigure2.AusergeneratesabusinessgraphusinganauthoringtooloragraphtoollikeMSExcel.Atanytimetheusermayactivatethecritiquingsystemforreviewingthegraphs/hedesignedsofar.Valuesofexternalfactslikethetypeofthechart,apieslice’scolororabar’slengtharetransmittedtothecritiquingsystem.Externalfactsandtheirrelationsto

Figure3.ArtifactHierarchy.

Theartifacthierarchyguaranteesthatpossiblecritiquesareshowninameaningfulway.Beforecritiquingapieslice’scolor,itmustbeclearthatapiechartistherightchoicetorepresentthedatatobevisualized.Whendevelopingacri-tiquingsystematfirsttheartifacthierarchyhastobesetup.Normallyafirsthierarchyisgivenbythedesigndomainitselfandbytheauthoringtoolusedtodevelopaparticu-lardesign.Modificationsofandextensionstothehierar-chymightbefuturenegotiationissues.Expertsmayrealizemodificationswiththehelpofanadministrator.Theartifacthierarchyautomaticallydeliversallpossibleexternalfactsforwhichconditionsmaybeformulated.Eachnodeoftheartifacthierarchytreerepresentsanexternalfact.

Filters.Conditionsforexternalfactscanbeformulatedusingso–calledfilters.Generatingthesefiltersrequiresknowledgeoftheunderlyingsystemandprogrammingskillswearenotdemandingoftheexperts.AccordingtoMackinlay[15]awidevarietyofdesignscanbesystemat-icallygenerated,ifcertainrequirementsarefulfilled.Thisimpliesthatapredefinedsetofbasicfiltersoratleastacom-positionofbasicfiltersissufficientforcompletelychecking

Figure2.ExampleforaCritiquingSystem.

0-7695-0493-0/00 $10.00 (c) 2000 IEEE3

Proceedings of the 33rd Hawaii International Conference on System Sciences - 2000

agivendesign.Asetofbasicfilterscanbedevelopedbyadministratorswhodonothavetobefamiliarwithapartic-ulardesigndomain.Basedonthesefiltersexpertsareabletobuildmacrofiltersonamoreabstractlevel.

Tosummarizetheprerequisites,inordertoprovideade-velopmentenvironmentforcritiquingsystemsanartifacthierarchywhichautomaticallyincludespossibleexternalfacts,andasetofbasicfiltersarerequired.Globalexpertknowledgeisdomain–specificandrepre-sentsuniversallyapplicableknowledge.Afurthertypeofdomain–specificexpertknowledgehasbeenidentifiedwhichdependsonaparticularparameter.

Expertknowledgemaydifferaccordingtoparameterslikeethnic,religion,orcountry,thereforeitiscalledlocalexpertknowledge.Withinthescopeofsuchaparametertheremayexistseveralshapingsoflocalexpertknowledge.

4.CompilingKnowledge

Theknowledgebaseofacritiquingsystemcontainsallinformationnecessaryforreviewingauser’sdesign.Inor-dertogetclearadvicetheknowledgebasehastobecon-sistentfromauser’spointofview.Ourunderstandingofconsistencyandinconsistenciesaredefinedinthefollow-ingsubsection.

4.1.ConsistencyandInconsistencies

Consistency.Theknowledgebasewillbeconsistentforusersofcritiquingsystems,ifthesystemdoesnotgiveanycontradictoryorredundantadvice.Nevertheless,contradic-tionscannotbeavoidedingeneral.Thissituationisex-plainedinParagraph’InconsistencyAnomalies’.Addition-ally,consistencyfromtheuser’spointofviewdoesnotim-plythattheentiredomainisalreadydepictedbytherulebase.

Inconsistencies.Ifadviceappearstobecontradictoryorredundant,inconsistencieswithintheknowledgebaseexist.Inconsistenciescanbeclassifiedintofourcategories:

Figure4.StructureoftheKnowledgeBase.

identity.Theruletobeaddedhasalreadybeendefinedbysomeoneelse.

refinement.Partsofarulearecoveredbyanotherrule.cycle/completeness.Acritiqueisalwaysshowntotheuserindependentofthevaluesoftheexternalfacts.conflict.Theruletobeaddedisincorrectinitself.

InconsistencyAnomalies.Withinpartsoftheknowledgebaseinconsistenciesshouldbepermitted.Thisiswhatwecallinconsistencyanomalies.Wehavetodistinguishbe-tweendifferenttypesofknowledge,theseare

globalexpertknowledgewhichisdomain–specificanduniversallyapplicable,

localexpertknowledgewhichisalsodomain–specific,butonlyvalidwithinitsscope,and

knowledgeofuserswhichisexpressedthroughausermodel.

TheenlargedrectangleinFigure4illustrateshowcountry–specificknowledgecanbedividedintoseveralpartitions.Thefollowingexampleistakenfrombusinessgraphdesign:Auserwantstocreateabusinessgraphforrepresentingthenumberofcinemagoerspermonth.Theuserchoosesabarcharttoillustratethedata.AccordingtoAgoston[1]peoplefromEuropepreferlesssaturatedcolorsthanpeoplefromtheUSA.Thereforetwoconflictingrulesareneededforcheckingthesaturationofthebars’colors:onerulebe-longstothe’Europe’partitionandtheotheronebelongstothe’USA’partition.However,rulesofbothpartitionsarevalid,sothatinconsistencieshavetobepermittedbetweenpairsoflocalexpertknowledge.Itisveryimportantthatusersofcritiquingsystemsaremadeawareofthisconflict-ingsituation.Theymustbeabletochoosebetweenonlyac-tivatingonepartitionofaparticularlocalexpertknowledgeoractivatingallpartitions.Ifonlyonepartitionisactivated,theknowledgebaseasawholewillbeconsistent.Ifmorethanonepartitionisactivated,inconsistenciesmayarise.Insuchsituationsthecritiquingsystemisnotabletodecidewhetheracritiqueisappropriateornot.Inourexample,itwouldnotbeclearwhethertoproposehighorlowsaturatedcolors.Thedecisionmustbelefttotheuser.However,ex-pertshavetoprovideseveralpartitions,sothatusershavethepossibilitytomakeadecisionatall.

Afurthertypeofknowledgeisrepresentedbyknowl-edgeofauser.Itisprovidedbyausermodel.User–dependentknowledgemaycontainpersonalpreferencesorindividualdisabilitiesofauser,company–dependentin-

0-7695-0493-0/00 $10.00 (c) 2000 IEEE4

Proceedings of the 33rd Hawaii International Conference on System Sciences - 2000

structionsorconcepts,etc.FurtherinformationisgivenbyOppermannin[16].

4.2.Objectives

Whenreviewingexistingcritiquingsystemswerealizedthatexpertsneedalotofsupportforcompilingtheirknowl-edgeefficiently.Fromthiswederivedthemainobjectivesofthiswork.Weidentifiedthefollowingissuestobeessen-tialforprovidingausefuldevelopmentenvironment.

specifyingknowledgewithasuitablelanguage.Thelanguagehastoabstainfromtheunderlyingmoretech-nicalspecificationofknowledgeorrules.Itshouldal-lowexpertsnotbeingfamiliarwithprogrammingtospecifytheirknowledge.Simulationandanalysisfa-cilitiesforthelanguagehavetobeprovided.supportingnegotiationofknowledge.Expertshavetobesupportedindiscussingknowledgeandagree-inguponparticularpartsoftheknowledgebaseinor-dertoimproveitsquality.Negotiationsareintendedtobebasedonvisualrepresentationsofknowledgeandpotentialinconsistencies.Thesevisualrepresen-tationsareintegratedinthelanguageusedforspecify-ingknowledge.Againsimulationfacilitiesareneededtohelpexpertstofindthebestandaboveallacorrectspecificationofparticularknowledge.

ensuringconsistencyoftheknowledgebase.Whenalteringtheknowledgebaseithastobeguaranteedthattheknowledgebaseremainsconsistentfromtheuser’spointofview.Newlyaddedormodifiedknowledgeischeckedwithalreadyspecifiedknowledge.

Fromtheseobjectiveswededucedageneralprocessmodelforcompilingknowledgewhichisdescribedinthefollow-ingsubsection.

Formulationofknowledge.Expertiseandknowledgeneedstobeformulatedinastandardizedway.Languageandmethodsusedforformulatingorspecifyingknowledgehavetobeapplicableforknowledgeofvariousfieldsandhavetomeetstrategiesofexpertsinvariousdomains.

Evaluationofknowledge.Evaluationhastoguaranteethatdifferentspecificationsofthesamefactarenotcontradic-tory.

Negotiationofknowledge.Negotiationissuesaregivenbyformulationproblemsandevaluationresults.ConcerningnegotiationwewouldliketofollowadefinitiongivenbyEasterbrook[5].Negotiationisdescribedasacollabora-tiveapproachtoresolvingconflictbyexhaustingallpos-sibilitiessothatanutmostsatisfactionisreachedforallparties.Itisoftencalledconstructivenegotiationinordertodistinguishitfromdistributiveorcompetitivenegotia-tion.AuthorsasDeBono[2]prefernegotiationtobere-strictedtoitsdistributivequality,includingaprocessofbid-dingandconcession–making.Easterbrookpreferstocalltheconcession–makingprocessbargaining,whichusuallyresultsinacompromise,whereastrueconstructivenegotia-tionseekstodevelopanewsolutionwhichfullysatisfiesallparticipants.

Distributivenegotiationmeansthatanexpertmoreorlesspersistsinimposinghis/herwill.Inourcaseitisveryimportanttomaketheassumptionthatallexpertsbeingin-volvedinanegotiationprocessareinterestedinbuildingaknowledgebasecooperatively.Thiscanonlyberealizedbyconstructivenegotiation.

Verificationofknowledge.Formulatedknowledgecanbeverified.Aformulationofknowledgemaystandincontradictiontoanotherformulationwhichbecomesvisible.Verificationallowstheanalysisofspecificationsofknowledgeandevaluationresults.

Thismodelgivesageneraloverviewoftheprocessesbeinginvolvedinexpertscompilingtheirknowledge.Re-lationshipsbetweentheseprocessesareindicatedbyarrowswhichshowtheflowofdata.Thetypeofdataisspecifiedinmoredetailwhenprojectingthismodelonatechnicallevel(seeSubsection4.4).Onceknowledgehasbeenformulated,itisnegotiated(6)orevaluated(1).Ifinconsistenciesoccur,evaluationresultshavetobeverified(2).Thismayresultinalteringtheformulationofknowledge(3)orinnegotiatingknowledge(4).Negotiationmaybesupportedbyverifica-tions(4)andmayleadtoalteringtheformulation(6)ormaytriggertheevaluationprocessagain(5).

4.3.AProcessModelForCompilingKnowledge

Theprocessmodelforcompilingknowledgeconsistsoffourmainprocessesformulation,verification,negotiation,andevaluation(seeFigure5).Thesearedefinedasfollows.

4.4.TechnicalLevelofProcessModel

Nowthegeneralprocessmodelisprojectedontoamoretechnicallevel.Abstractdescriptionsofprocessesarere-placedbyconcreteonesandinterdependenciesbetweensin-

Figure5.ProcessModelforCompilingKnowledge.

0-7695-0493-0/00 $10.00 (c) 2000 IEEE5

Proceedings of the 33rd Hawaii International Conference on System Sciences - 2000

gleprocessesareexplainedinmoredetail(seeFigure6).Italsodemonstratesascenarioforinteractionstakingplacewhenexpertscompiletheirknowledgeforbuildingaknowl-edgebase.Fromthefollowingscenariowederivethenego-tiationissuesbeingrelevantforcritiquingsystems.

5.NegotiationSupport

Asdiscussingknowledgeistheonlywayofbuildingacommonviewofadomain,negotiationsplayanimportantrolewhencompilingknowledgecollaboratively.

5.1.NegotiationIssues

Fromabovewecanmainlyidentifythreesituationswhereexpertsaredependentonnegotiatingknowledge.Thesesituationsoccurwhen:extendingtheartifacthier-archy,askingforadvicewhenformulatingaruleanddis-cussingandeliminatinginconsistencies.Thus,thefollow-ingnegotiationissuescanbeidentified.

Figure6.TechnicalLevelofProcessModel.

Astartingpointmaybeanexpertspecifyinghis/herknowl-edge.Asknowledgeisrepresentedbyrules,specifyingknowledgeisequivalenttoeditingrules.Whentryingtomodifytheknowledgebase,i.e.,addingormodifyingarule,aconsistencycheckismadefirst.Deletingarulecanbeignoredbecausefromtheuser’spointofviewitcannotcauseanyinconsistencies.Iftheknowledgebaseremainsconsistent,anewrulewillbeadded,oranalreadyexistingonewillbemodified.Ifinconsistenciesoccur,expertswillbeabletodebugandalsotoanalyzeinconsistencieswithaffectedrules.Afterdetectingthereasonsforinconsisten-cies,theexpertmaygoovertheruleagainandeliminatetheinconsistencieswithhelpoftheeditor.Inordernottomodifyorremoveanexistingruleinfavorofhis/herownrule,expertsshouldcontactauthorsofaffectedrules.Theymaystarttonegotiatepurposeandcompositionofrules.Theyhavetoagreeonwhethertomodifyorremoverulesorwhethertomergerules.Expertsmayalsodecidetore-moverulesalreadyresidingintheknowledgebase.Theexistenceofruleswithinthedatabasemeansthatthereexistnoinconsistencieswithotherrules.Thisdoesnotguaranteethataruleismeaningfulatall.Constructingknowledge–adding,deleting,modifyingofrules–iscoupledwithau-tomaticallygeneratingarecordofanaction.Asthereareseveralexpertsconcernedwithmaintainingtheknowledgebase,itisveryimportanttomakeallmodificationstranspar-enttoallexperts.Doingso,theyareabletokeeptrackofallmodificationstakingplace.Especiallyifexpertsarenotinvolvedforaperiodoftime,theknowledgebase’shistorygivesexpertstheopportunitytogetthroughalleventstheyareinterestedin.Expertswouldnotbeabletomakeanydecisionswithoutknowingallfacts,i.e.,modificationsoftheknowledgebaseduringtime.Automaticallygeneratedrecordscanbeextendedmanuallybyindividualcommentsandannotationsofexperts.

theartifacthierarchy.Afirstartifacthierarchyisgivenbythedesigndomainitselfandtheauthoringtoolusedtodevelopaparticulardesign.Modificationsofandextensionstothehierarchyarenegotiationis-sues.Aseachnodeoftheartifacthierarchytreerepre-sentsanexternalfact,extensionstothehierarchyallowexpertstospecifyrulesforfurtherexternalfacts.ThecoherencebetweenartifacthierarchyandexternalfactsisdescribedinSection3.

arule’scomposition.Expertsmaydiscussarule’sconditions,critiquetext,andexternalfacts.Theover-allpurposeofaruleisalsoanissue.

inconsistencies.Ifinconsistenciesoccurwhentryingtoenhanceoraltertheknowledgebase,authorsorex-pertsofaffectedruleswillhavetonegotiatetherules’meanings.Further,theyhavetoagreeonwhethertocorrectorremovearuleortomergerules.

Currently,wefocusonnegotiatinginconsistencies,whichareanoutcomeoftheconsistencycheckprocess.

5.2.SupportingWell–StructuredNegotiations

Bywell–structurednegotiationsweunderstandthateachnegotiationfollowsacertainsequenceofactions.Thisisachievedbybasingthenegotiationprocessononeoftheex-istingnegotiationprocessmodels.Kerstendescribesin[13]ageneralprocessmodelfornegotiation,whichconsistsoffiveconsecutivephases,startingwiththeplanningofnego-tiationandclosingwithanachievedagreement.Kersten’smodeltakesthepresentandfutureuseofcomputingandcommunicationtechnologiesintoconsideration.Itisalsodesignedtosupportactivitiestypicalforgroupdecisions.GroupdecisionsupportandKersten’sintentiontosupportconstructivenegotiationarethemainreasonsforourdeci-siontobaseournegotiationprocessesonKersten’sprocessmodel.InthefollowingweintroduceKersten’snegotiationprocessmodelandpointouttherelevanceofeachphase

0-7695-0493-0/00 $10.00 (c) 2000 IEEE6

Proceedings of the 33rd Hawaii International Conference on System Sciences - 2000

withregardtonegotiatingarule’scompositionandincon-sistenciesbetweenrules.ForeachphasetheitalicizedtextdescribesKersten’sintention.Ourreflectionsandadapta-tionstonegotiationofknowledgealwaysfollow.

Phase0:Initiatingnegotiation

ThisphasedoesnotbelongtoKersten’sprocessmodel.ItisratheraphasewhichprecedesKersten’sfivephases.Phase0isintendedtoestablishafirstcontactbetweentheini-tiatorofanegotiationandnegotiatingparties.Normallythesearetheauthorofarulewhichcausedinconsistencieswithalreadyexistingonesandexpertsofaffectedrules.Aworkspaceiscreatedfortheongoingnegotiationandallpo-tentialparticipantsareinvited.

Phase1:Searchingforarena,selectionofcommunicationmode

ThefirstphaseofKersten’sprocessmodelfornegotiationdealswithsearchingforavenuewherenegotiationshouldtakeplace.

Thelocationiseitheraphysicaloravirtualsetting.Furtherthemodeofcommunicationhastobedetermined.Infor-mationisexchangedsynchronouslyorasynchronously.Ourdevelopmentenvironmentprovidesvirtualsettings,becauseitisintendedtosupportexpertsbeingdistributedallovertheworld.Consideringdifferenttimezonesbothmodesofcommunication,asynchronousaswellasanasynchronouscommunicationmode,arerelevant.Theselectedmeansofcommunicationrepresentthemediumforfulfillingtheac-tionpointsbeingsetinphase2tophase5.

Phase2:Agendasetting,determiningnegotiationissuesThesecondphaseconsistsofsettingtheagendafornegoti-ation.Thisincludesdiscussingandagreeingonthetermi-nologyandtheissuestobedecidedupon.

Terminology.Evenifexpertsworkontopicsbelong-ingtothesameareaofresearch,theymayuseadifferentterminology.Althoughthevocabularyispartlyfixedbytheartifacthierarchy,expertsmayeasilyfallbackontermstheyarefamiliarwith.Intheendacommonbasisisneededinordertoavoidmisunderstandings.Weintendtoprovidefastlook–upmechanismsenablingexpertstocheckunknownorambiguousexpressionsinacentralizedglossary.Thus,ex-pertscanrelyontheterminologytheyalreadyagreedonandtheydonothavetocometoanagreementanew.

NegotiationIssues.Theissuestobedecidedonaregivenbyarule’scomponentsorinconsistencies.Ifanex-pertinitiatesanegotiationofaruleduringtheconstructionphaseofthisrule,s/hedeterminestheruleorrule’scom-ponents/hewouldliketodiscuss.Ifanegotiationisini-tiatedbecauseofinconsistencies,thenegotiationissuesaregivenbytheresultsoftheconsistencycheck.Theauthorsoftheseruleshavetodecidewhethertokeeponeoftherulesorwhethertofindanewsolution,thatmeansspecifyinganewruleexpressingtheirknowledge.

Phase3:Exploringthefield

Thisphasemainlydealswithfurtherproblemspecificationanditsanalysis.

InordertoexplorethefieldnotonlythenegotiationissuesdeterminedinPhase2arediscussedbutalsoarule’spur-poseingeneral.Ithastobequestionedwhetheraruleex-pressesexactlywhatanexpertintended.Expertsareen-couragedtodiscussarule’scomposition.Firstly,theyhavetoagreeontheexternalfactsforwhicharuleisformulated.Thenextstepistodecideontherule’spreconditions.Fi-nally,thecritiquetextisdeterminedwhichisshowntotheuserifthevaluesoftheexternalfactsfulfilltheprecondi-tions.Thestepwiseexplorationofarule’scompositionissupportedbysimulations.Implicationsofrulemodifica-tionscanbetestedwithhelpofaconsistencycheck.Phase4:Narrowingthedifferencesandsearchforintegra-tion

Phase4ismainlymeantforrefiningobjectivesandlim-itationsthroughexchanginginformation.Theselectionandverificationofstrategiesarereconsideredandpartiesshouldrealizethepotentialofcompromiseandconsiderconcessions.

Basedonfindingsfromphase3,resultsofsimulationsanddebuggingsessionsareevaluated.Learningoftheothers’objectivesmayleadtoacompromise.AcompromisemayresultinseveralpartitionsoflocalexpertknowledgeasitisdescribedinSubsection4.1.

Phase5:Searchforagreementandimprovements

Criticalissues,thepartieswerenotabletodecideupon,areidentified.Lookingforcompromisesandjointproposalsandlooseninglimitationsarefurtheraspects.

Ifexpertsfailincomingtoanagreementinphase4,thisstagewillsupporttheidentificationofthecriticalissuesandareasofdisagreement.Thisenablesexpertstocometoaso-lutionjointly.ThelineofproceedingalreadyusedinPhase4isemployedhere.

6.ColDECS–COMPONENTS

AccordingtoourconceptsColDECSisdividedintofivemaincomponents(seeFigure7).Theyreflecttheprocessesidentifiedwithinourprocessmodelforcompilingknowl-edge(seeSubsection4.3).

Theserverprovidescommunicationchannelsbetweenconsistencycheckerandclients–representedbygraphicaluserinterfaces–anditalsohandlestheaccesstotheknowl-edgebase.Furthertheservermanagesjointviewsandcon-trolofconnectedclients.Thisincludesfloorcontrolpoliciesforregulatingaccesstosharedrules.

Thegraphicaluserinterfaceprovidesnavigationfacil-itiesforaccessingruleswithintheknowledgebase.Rulesareorganizedinprojects.Projectsholdinformationabouta

0-7695-0493-0/00 $10.00 (c) 2000 IEEE7

Proceedings of the 33rd Hawaii International Conference on System Sciences - 2000

serverabouttheresult.Resultsarepresentedtotheuserthroughthegraphicaluserinterface.

Theknowledgebasestoresrules.Wehavetodistin-guishbetweenrulesbeingmarkedasvalidorinvalid.Ruleswhichstillhavetobecheckedforconsistencywithalreadyexistingonesaremarkedasinvalid.Thiskeepstheknowl-edgebaseconsistentforusersofcritiquingsystems,whousethesystemwhileitsbeingmodified.Afterasuccessfulconsistencycheckrulesaremarkedasvalid.Nowtheycanbeaccessedforcritiquingauser’sdesign.

7.AFirstPrototype

AfirstprototypeofColDECSfocusesnegotiationsup-portanditisbasedonVIPspace(VIsuallyProgrammableSharedWorkspace)whichhasbeenintroducedbyDueckeretal.in[4].VIPspacehasbeendevelopedformanag-ingtheasynchronousexchangeofinformationinformofobjects,e.g.,documentsandimages.Itisadaptabletouser–specificneedsandtask–specificrequirementsviavisualprogramming.TheVIPspacedesktopisdividedintoaglobalworkspaceforsharingobjectsandaprivateworkspaceformanagingobjectsindividually.Ontheonehandweprovideaclearseparationbetweentheglobalandprivateworkspace.Ontheotherhandobjectscanbemovedfromtheprivatetotheglobalworkspaceveryeasilyusingahomogeneousdraganddropparadigm.Theglobalaswellastheprivateworkspaceholdtaskfieldswhichareeitherpredefinedorvisuallyprogrammable.Taskfieldsareusedforstructuringandprocessingobjectsandtheycanbepro-grammedbyspecifyingif–thenrulesusingicon–andform–basedvisualmeans.Rulesaredefinedbyconditionsandactions.Aruleisperformedwhendroppinganobjectonataskfield.Ifarule’sconditionisfulfilledthecorrespondingactionwillbeexecuted.

Furtherexistingcomponentsaretheconsistencycheckerandtheruleeditorwhichprovidesvisualmeansforspec-ifyingfirstorderlogicexpressions.Thisiscomparabletobuildinganetworkoflogicgates.Theeditoralsogivesac-cesstosimulationanddebuggingfacilities.

Figure7.ColDECSSystemComponents.

rule’sauthorincludinghis/hercontactaddress,thedateofcreationormodification,keywordsforenablingsearchesforaparticularrule,andrecordsandannotationsforkeep-ingtheknowledgebase’shistory.Anintegratedruleeditorandannotationsystemallowthemanipulationofrulesandruleannotations.Theruleeditorallowsthecompositionofrules.Externalfactsandfiltersarepredefined.Assumingthatandareexternalfactsandassumingthat,,andarefiltersanabstractrepresentationofacomposedrulemaylookasfollows

if

or

then

and

Negotiationsarehandledwithinanegotiationspace.In-volvedpartiesareinvitedbytheinitiator.Negotiationissuesareprocessedandrelevantinformationaboutthediscussionandpartialsolutionsarerecorded.Especiallywhenusinganasynchronouscommunicationmode,itisnecessarytokeeptrackofanongoingnegotiation.Resultsofaconsistencycheckaremadevisibletotheuser.Theycanbeanalyzedwithhelpofadebugger.Valuesofexternalfactscanbechosenasinputforaruleandfollowedstepwisewhenpass-ingsinglefilters.

Thebrowserprovidesspecialsearchfacilitiesandpre-sentationsofsearchresults.Beforespecifyinganewruleexpertsmaysearchforrulesalreadyformulatingtheex-pert’sknowledge.Dependingonkeywords,rulesfulfill-ingthesesearchcriteriaareshowntotheuser.Relationsbetweensearchcriteriaandruleswillbevisualized.Thelengthoftheconnectionbetweenrule(documentsymbol)andsearchcriteria(cycle)isameasurementforsimilarity.Theconsistencycheckeristriggeredbytheserver.Theconsistencycheckergetstheruletobecheckedandevalu-atesitscomposition.Theconsistencycheckerinformsthe

7.1.EnhancementsforSupportingNegotiation

ForthepurposeofsupportingnegotiationsVIPspacehasbeenenhancedbysynchronousCSCWtools.Theseareatextconferencingtoolandasharedwhiteboard.Wheneverthesetoolsarestarted,thetools’ownergetsadedicatedcolorformakingcomments.Itisthesameforwhiteboardandtextconferencingsothatafastandeasyassignmentofcolortoexpertispossible.

SharedWhiteboard.Thewhiteboarddisplaysthenego-tiationissuewhichhasbeenprocessedalready.Anexpert’snewlyspecifiedrule,theinconsistency,andaffectedrulesarevisualizedpairwise.

0-7695-0493-0/00 $10.00 (c) 2000 IEEE8

Proceedings of the 33rd Hawaii International Conference on System Sciences - 2000

Textconferencing.Asthevisibleareaofthewhiteboardislimited,thetextconferencingtoolisusedforformulatingmorecomplexcontributionstoanegotiation.Textconfer-encingenablesexpertstodiscusstheoverallpurposeandcompositionofrules.

Application–specificObjects.Someapplication–specificobjects,e.g.,forrulesandinconsistencieshavebeendefinedinordertosetupVIPspace.Taskfieldshavebeenprogrammedforstructuringobjectsoftyperuleandinconsistency.ThetaskfieldConsistencyCheckhasbeenprogrammedforlaunchingtheconsistencychecker.

Theseenhancementsallowexpertstostructuretheirrules,toexchangerulesandtodiscussarule’scomposi-tionandinconsistenciesasitisdescribedinthefollowingscenario.

arepairwisedistinguishablebecomestrue,inparticularcol-orsofpieslicesbeingneighboredaredistinguishable.Thetextconferencingtoolallowsexpertstotalkabouttherules’compositions.Bydoubleclickingontheobjectoftypein-consistencytheinitiatoradditionallyrequestsparticipationfromallexpertson–line.Expertsjointheconferencebysimplyconfirmingtheinitiatorsrequest.Allconferencingtoolsarestartedautomatically.

8.ConclusionsandFutureWork

Wehaveintroducedconceptsforacollaborativede-velopmentenvironmentfordesign–orientedcritiquingsys-tems.ColDECSsupportsexpertsinconstructingaknowl-edgebasewhichrepresentsanessentialpartofcritiquingsystems.Aprocessmodelforcompilingknowledgehasbeendevelopedinordertobuildthebasisforthesystem’sarchitecture.Integratingexpertsfromdifferentresearchfieldsinbuildingaknowledgebaserequiresnegotiationsupport.Weidentifiednegotiationissuesarisingwhencom-pilingknowledgecollaboratively.Sofarnegotiationissues,inparticularinconsistencies,areprocessedwiththehelpofasharedwhiteboard.Thisjointviewispartofafirstproto-typewhichhasbeensetupbaseduponVIPspace.Nextstepismakingtheruleeditorcapableofmulti–useractionssothatinconsistenciescanbevisualizedbytheeditorandrulescandirectlybemanipulated.ResultsofpreliminaryusertestsconcerningtheusabilityofVIPspaceareverypromis-ing.Furtherusabilitytestsexamininghowexpertshandlenegotiationusingthesharedwhiteboardandtextconferenc-ingwillgiveusefulindicationsofhowtoputthisintoprac-tice.Currently,theruleeditorcanonlybeinvokedfromanexpert’sprivateworkspace.Enhancingtheruleeditorad-ditionallymakestheintegratedsimulationanddebuggingfacilitiesaccessibleforallexperts.

7.2.UsageScenario

Thisscenariodescribestwoconsecutiveprocesses.Thefirstonepointsouthowtocometoasituationwherenego-tiationbecomesnecessary.Thesuccessiveprocessshowshowanegotiationisperformed.Figure8showsascreen-shotofVIPspaceandillustratesbothprocesses.Examplesrefertotheareaofbusinessgraphdesign.

Howtocometoasituationwherenegotiationbecomesnecessary.TheprivateworkspaceofVIPspaceisusedforpreparingaruleforpublishing.Doubleclickingonaruleobject,e.g.,PieColorinthetaskfieldMyRulesopenstheruleeditorforthisobject(1).TherulePieColorisde-finedforindicatingthatcolorsofdifferentpieslicesofapiechartarenotdistinguishable.TheexpertmaydecidetopublishtherulePieColor.ThisisdonebymovingthepreparedruletotheglobaltaskfieldConsistencyCheckviadraganddrop(2).Thistaskfieldisprogrammedtoevalu-atethedroppedobjectoftyperuleindependenceonalreadyexistingrules.Ifinconsistenciesoccurtheywillbeautomat-icallydisplayedintheglobaltaskfieldInconsistencies(3).Howtoperformnegotiation.Anegotiationstartswiththepreparationphase,inwhichtheinitiatorofanegotia-tioncontactstheauthorsofaffectedrules.Establishinganinitialcontact,e.g.,byE–mail,mayleadtoanimmedi-ateconferenceoranappointmentforaconference.AtthetimeagreedallexpertslogonVIPspaceforattendingthenegotiation.Theinitiatorinvokestheconferencingtools;theseareasharedwhiteboard(4)andatextconferencingtool(5)bydoubleclickingonanobjectoftypeinconsis-tencyinthetaskfieldInconsistencies.Thesharedwhite-boardisopenedshowingthenewlyspecifiedrulePieColorandtherulePieNeighborwhichcausedtheinconsistency.Locationandtypeoftheinconsistencyrefinementinthiscasearevisualized.Iftheconditionthatcolorsofpieslices

References

[1]G.A.Agoston.ColorTheoryandItsApplication,Artand

Design.SpringerVerlag,2edition,1987.[2]E.D.Bono.Conflicts.PenguinBooks,1985.

[3]T.X.BuiandM.F.Shakun.Negotiationprocesses,evolu-tionarysystemsdesign,andnegotiator.GroupDecisionandNegotiation,5,1996.

[4]M.Duecker,W.Mueller,andJ.Rubart.Innovativecon-ceptsforconfiguratingsharedworkspacesthroughvisualprogramming.InProceedingsoftheThirty-SecondAnnualHawaiiInternationalConferenceonSystemSciences1999,Hawaii,USA,January1999.

[5]S.Easterbrook.Handlingconflictbetweendomainde-scriptionswithcomputer–supportednegotiation.KnowledgeAquisition:AnInternationalJournal,3:255–2,1991.[6]G.FischerandA.Girgensohn.End-usermodifiabilityinde-signenvironments.InHumanFactorsinComputerSystems,CHI’90.ACM,1990.

0-7695-0493-0/00 $10.00 (c) 2000 IEEE9

Proceedings of the 33rd Hawaii International Conference on System Sciences - 2000

Figure8.AFirstPrototype.

[7]G.Fischer,A.C.Lemke,T.Mastaglio,andA.Morch.The

roleofcritiquingincooperativeproblemsolving.ACMTransactionsonInformationSystems,9(2):123–151,1991.[8]P.H.Gulliver.DisputesandNegotiations.AcademicPress,

NewYork,1979.

[9]B.Gutkauf,S.Thies,andG.Domik.Auser-adaptivechart

editingsystembasedonusermodelingandcritiquing.InUserModeling:ProceedingsoftheSixthInternationalCon-ferenceUM97.SpringerVerlag,1997.[10]R.P.H¨am¨al¨ainenandE.Kettunen.On–linegroupdecision

supportbyhipre3+grouplink.InProceedingsoftheThirdInternationalSymposiumontheAnalyticHierarchyProcess,WashingtonD.C.,pages7–557,July11–131994.

[11]B.Harstad.NewApproachesforCritiquingSystems:Plu-ralisticCritiquing,ConsistencyCritiquing,andMultipleIn-terventionStrategies.PhDthesis,DepartmentofCom-puterScience,UniversityofColoradoatBoulder,December1993.

[12]C.HolsappleandA.Whinston.DecisionSupportSystems–

AKnowledge–BasedApproach.WestPublishingCompany,1996.

[13]G.E.KerstenandD.Cray.Perspectivesonrepresentation

andanalysisofnegotiations:Towardscognitivesupportsys-tems.GroupDecisionandNegotiation,5(4–6):433–469,1996.

G.E.KerstenandS.Noronha.Negotiationviatheworldwideweb:Across-culturalstudyofdecisionmaking.GroupDecisionandNegotiation,(toappear).

J.Mackinlay.Automatingthedesignofgraphicalpresen-tationsofrelationalinformation.ACMTransactionsonGraphics,5(2):110–141,1986.

R.Oppermann,editor.AdaptiveUserSupport.LawrenceErlbaumAssociates,Hillsdale,NewJersey,1994.

B.G.Silverman.Surveyofexpertcritiquingsystems:Prac-ticalandtheoreticalfrontiers.CommunicationsoftheACM,35(4):106–127,April1992.

G.Stahl.InterpretationinDesign:TheProblemofTacitandExplicitUnderstandinginComputerSupportofCooperativeDesign.PhDthesis,DepartmentofComputerScience,Uni-versityofColoradoatBoulder,November1993.

Y.Yuan,J.B.Rose,N.Archer,andM.Suarga.Aweb–basednegotiationsupportsystem.InB.F.Schmid,D.Selz,andR.Sing,editors,EM–ElectronicContract-ing.EM–ElectronicMarket,number3in8.Septem-http://www.electronicmarkets.org/-ber1998.URL:

netacademy/publications.nsf/all

[14][15][16][17][18]

[19]

0-7695-0493-0/00 $10.00 (c) 2000 IEEE10