Understanding Group Decision-Making

ANewDirectioninGroupSupportSystemTheory:AnEvolutionaryApproachtoGroupDecision-MakingJackieRees1andGaryJ.Koehler21PurdueUniversityKrannertGraduateSchoolofManagementWestLafayette,IN47907Phone:[phone]Fax:[e-mail],2UniversityofFloridaDepartmentofDecisionandInformationSciences351StuzinHallWarringtonCollegeofBusinessGainesville,FL32611Phone:[phone]Fax:[e-mail]2ANewDirectioninGroupSupportSystemTheory:AnEvolutionaryApproachtoUnderstandingGroupDecision-MakingAbstractWeproposemodelingGroupSupportSystem(GSS)searchtaskswithGeneticAlgorithms.UsingexplicitmathematicalmodelsforGeneticAlgorithms,(GAs),weshowhowtoestimatetheunderlyingGAparametersfromanobservedGSSsolutionpath.Oncetheseparametersareestimated,theymayberelatedtoGSSvariablessuchasgroupcompositionandmembership,leadershippresence,thespecificGSStoolsavailable,incentivestructure,andorganizationalculture.TheestimatedGeneticAlgorithmparameterscanbeusedwiththemathematicalmodelsforGAstocomputeorsimulateexpectedGSSprocessoutcomes.SubjectAreas:GroupDecisionSupportSystems(GDSS),GeneticAlgorithms,ComputationalModelingandMarkovModels.1.IntroductionWorkperformedbygroupsplaysanimportantroleinthesuccessorfailureoftoday’sorganizations.Technologyisoftenusedtosupportthetasksundertakenbythesegroups.GroupSupportSystems(GSS)encompassingGroupDecisionSupportSystems(GDSS)andElectronicMeetingSystems(EMS)areofparticularinterest.Muchattentionhasbeenfocusedongrouptaskoutcomesandtheimpactoftechnology.However,thereisstillmuchtobelearnedabouthowgroupswillperformgivencertainenvironments,tasksandgroupmembership.OrganizationsutilizeGroupSupportSystemsinordertoimprovetheeventualoutcomesofgroupmeetings.GiventheinterestintheeventualoutcomeofGSSuse,theprocessesleadingtothesetofoutcomesneedtobecarefullyexamined.Groupsusingthesetoolsmostoftenhaveaspecifictaskorproblemtoaddress.Inordertoaddresssuchproblems,groupsmustexplorea3spaceofpossiblesolutions.Themovementthroughthesearchspacecanbeviewedastheideagenerationorbrainstormingprocess.Themovementfromsolutiontosolutionisachievedthroughcreativeinsights,negotiationandgrouplearning.Asgroupmembersexchangeinformation,newsolutionsarediscovered,potentiallybetterthanprevioussolutions.Thegroupadaptsitssearchaccordingtoseveralfactorsincluding,butnotlimitedto,thegroupcompositionandmembership,leadershippresence,thespecificGSStoolsavailable,incentivestructure,organizationalcultureandmostimportantlyinputandfeedbackfromgroupmembersasthesearchprogresses.Occasionally,acompletelynewlineofthinkingisundertakenorarandomideaisinsertedasapotentialsolutioninthesearchspace.Wesuggestthatideas,proposalsandsolutionsfromgroupsusingGSScanbeviewedasstringsofgenesinanevolving,adaptiveenvironment.Theideasofthegroupevolveuntilasolutionorsetofsolutionsisreached.Thisisnotanewidea-Hirokawaand.Johnson(1989)arguedsimilarly.Eventhe“randomideafromleft-field”fitsthisviewasmerelyamanifestationofpunctuatedequilibria–ahallmarkofevolutionarymethods.However,webuildonthisobservationbyproposinganexplicitmodelofevolutionarybehavior–theprocesscapturedbythesimpleGeneticAlgorithm.ThereareseveralusefulmodelsofGSSavailablewhichprovideinsightintoparticularGSSprocessesandpotentialoutcomes.Severalofthesemodelsaredescriptive(PooleandDeSanctis,1990;Nunamakeretal.,1991;Hiltz,1988;RaoandJarvenpaa,1991),andothersaremoreanalytical(GavishandKalvenes,1996;ValacichandDennis,1994).TheanalyticalmodelforGSSweproposedepartsfromotherGSSmodelsinthatitprovides,throughanalogy,acomputationalviewofGSSprocessesandexpectedoutcomes.TheanalogyutilizedisthesimpleGeneticAlgorithm(GA).4TheuseofgeneticalgorithmsasthebasisofamodelforGSShasseveraladvantagesandimplications.Forexample,overthelastseveralyears,asoundmathematicaltheoryhasbeendevelopedthatdescribestheexactexpectedbehaviorofGAs.Thistheory,inprinciple,couldbeusedtodeterminemanyGSScharacteristics,suchasexpectedtimetillanoptimalideaisgenerated.Also,thereisawidebodyofpractitionerrules-of-thumbforGAsthatcouldproveusefulindesigningbetterGSSprocesses.Theremainderofthispaperisorganizedasfollows.Section2presentsrelevantbackgroundonGroupSupportSystemsandGeneticAlgorithms.InSection3wesummarizeourargumentsforrepresentingGSSbrainstormingandideagenerationasaGeneticAlgorithm.InSection4wediscussvariousspecificsofGAimplementations.InSection5weshowhowtoestimatetheunderlyingGAparametersfromGSSsolutionpaths.ThisrequiresapresentationoftheunderlyingmathematicalmodelsforGAs.InSection6wediscussspecificissuespertainingtomodelingaGSSasaGA.InSection7wepresentresultsfoundwhenestimatingGAparametersfromactualGSSdatausingtheestimationprocedureofSection5.Section8showshowaGAmodelcanbeusedtoderiveprocessvaluessuchastheexpectedtimetoseeanoptimalsolution.Finally,inSection9weprovideourconclusionsandfuturedirections.2.BackgroundThefollowingsub-sectionsproviderelevantbackgroundonthevariousresearchareasthatformthebasisforthisresearch.Section2.1providesabriefsurveyoftheoreticalandcomputationaldevelopmentsinGSS.Thissub-sectionhighlightstheneedforadditionalinvestigationintoc