Submitted to Adaptive Behavior (51598) (resubmitte

CollectiveProblemSolving:FunctionalitybeyondtheIndividualNormanL.JohnsonTheoretica[e-mail][phone],[phone]faxSubmittedtoAdaptiveBehavior(5/15/98)(resubmitted11/5/98)SpecialIssueonSimulationofSocialAgentsKerstinDautenhahn,EditorCollectiveProblemSolvingAbstractFollowinganon-reductionistapproachtotheexplanationofhigherfunctionalityobservedincollectiveproblemsolvers,asimpleagent-basedmodelisusedto“solve”asequentialproblem-amaze.Largercollectivesoftheindividualagentsareobservedinthesimulationstolocateaminimalpath,eventhoughtheagentsarenon-interacting,havenoglobalperceptionofthemazeanduserulesthatdonotincludelogicforfindingashorterpath.Theconvergencetoanoptimalpathisarguedtobeademonstrationofbothanemergentproblemformulationandemergentproblemsolution.Furthermore,manyofthedynamicsandpropertiesofcooperatingcollectivesareobserved:performanceofthecollectivegreaterthanthatoftheaverageindividual,reducedperformancewithlessdiversity,abilitytofunctioninthepresenceofextremenoiseandinformationloss,improvedcollectiveperformancewithestablishedindividualproblemsolvers,pathsensitivitytoindividualcontributionsbutlimitedsensitivityofgroupperformance,andothers.Theimplicationsoftheresultstotheformationofself-organizingknowledgeanddecisionmakingsystemsarediscussed.Keywords:diversity,collective,self-organizing,emergentproblemsolvingNormanL.Johnson“solving”theirown“problems”resultincollectiveproblemdefinition,knowledgecreationorproblemsolvingwhichisgreaterthanthatoftheindividual?Byindividuals,wemeanorganisms,groups,organizations-anyentitythatislocalizedinphysicalorconceptualspace.Byproblems,wemeanbothproblemsthatareconsciouslydefinedandonesthatarenotexplicitlystatedbutarestillsolvedbythelawsorstructuresgoverningtheglobalsystem.Atypicalapproachtoansweringthisquestionmightbetheexaminationofasystemwithinteractingagentsthatcooperateandcompetewithsomeknowledgeofsharedresources,beitecologicalniche,commoditymarket,orsocialposition.Theindividualsareoftentakentobeagentswhocanmodifytheirbehaviorbasedupontheirassessmentoftheirrolesandoutcomes.Asresearchersweoftenattributesignificantcapabilitytoouridealizedagentsinordertoexplaintheobservedfunctionalityatagloballevel.Butwhatifthereareothermechanismsforfunctionalitythatarebeingoverlooked.Tworealworldexamplesaregiventomotivatethepossibilityofanalternativeviewpoint.Thefirstexampleistheformationofwalkwaysfollowinganewbuildingdevelopment-amodern-dayexampleofpathformationinnature.Thepre-determinationofwalkwayswhichbestcapturespathpreferencesoftheusersisoftenanexerciseinfolly,asjudgedfromthealternativepathwaysthatpeoplequicklydevelop.Theseplannedsolutionsprobablyfailbecauseofthemultiplicityofthefactorstobeconsidered:differentdestinations,terrain,security,exposuretotheweatherandmodesoftravel.Someplannershavelearnedthatoftenthebestsolutionistoletthe“system”determinethepathsbyfirsthavinggrasswithnopathsandthengraduallyconvertingemergingpathstoformalwalkways.Thesefinalpathsrepresentthecollectiveactionofmanyindividualssolvingtheirownpathproblem,inamannerthatisultimatelyusefultotheentirepopulationbutwhichisneverexpressedasagoalattheleveloftheindividual.AsecondexampleistherecommendedbooklistsatAmazon.com,anonlinebookstore.Whenacustomerfindsapossiblebookofinterest,thereisalsoshownalistofbooksthatarerelated.Theselistsareconstructedbydisplaying,accordingtofrequency,thebooksthatwerepurchasedbypeoplethatalsopurchasedthefoundbook.Thelistsareexceptionallyusefultosearchasequenceofrelatedbooksuntiladesiredbookisfound.Giventhatthepossiblechoicesexceedamillionbooks,staffsofhumanbooksellerswouldhavegreatdifficultywiththesuccessratesofthisrecommendationmethod.Yet,thiscapabilityisfoundedonthesimpleprocessofcapturingthepurchasehabitsofindividuals.Theseexamplesillustratehowcollectivefunctionalityatagloballevelcanoccurwithoutintentionalproblemsolvingonthepartoftheindividual.Arguablybothoftheseexamplesinvolvemechanismsthatreinforceemergingpatterns(selectionofanexistingpathorbook),butequallyarguablybothexamplescanexhibitcollectivefunctionalityevenifthesepositivereinforcingmechanismsareeliminated.Howthenisitpossibletoreconcilethetraditionalapproachesofcollectiveproblemsolvinginvolvingcooperationandcompetitionofglobally“aware”individualsandtheaboveexamplesofglobalproblemsbeingsolvedwithoutawarenessoftheindividual?Thereisagrowingbodyofliteraturethatisaddressingthisreconciliationinthefieldsofbiology,economicsandsociology.ApivotalanalysisoftraditionalapproachestocollectiveactionbyevolutionarybiologistsandcognitivescientistswaspresentedbyHemelrijk(1997).Throughastraight-forwardsimulationofherdformationthatincludesonlyaggressiveagents(i.e.,thereisnoinherentmechanismforcooperationembodiedintheindividual),thecollectivebehaviorofcooperationisobserved.Sheconcludesthatcooperativebehaviorfromessentiallyuncooperative