We reformulate the agent-based opinion dynamics models of Weisbuch-Deffuant and Hegselmann-Krause as interactive Markov chains. So we switch the scope from a finite number of n agents to a finite number of n opinion classes. Thus, we will look at an infini
arXiv:0708.3293v1 [physics.soc-ph] 24 Aug 2007
ContinuousOpinionDynamics:InsightsthroughInteractive
MarkovChains
JanLorenz
DepartmentofMathematicsUniversityofBremen
Nienburgerstr.42,28205Bremen,Germany
email:math@janlo.de
InProceedingsofIASTEDConference”Modelling,SimulationandOptimizationMSO2005”,
August29-31,2005,Oranjestad,Aruba
ABSTRACT
Wereformulatetheagent-basedopiniondynamicsmod-elsofWeisbuch-Deffuant[1,2]andHegselmann-KrauseasinteractiveMarkovchains.Soweswitchthescopefroma nitenumberofnagentstoa nitenumberofnopinionclasses.Thus,wewilllookatanin nitepopula-tiondistributedtoopinionclassesinsteadofagentswithrealnumberopinions.
TheinteractiveMarkovchainsshowsimilardynami-calbehaviorastheagent-basedmodels:stabilizationandclustering.Ourframeworkleadstoa’discrete’bifurcationdiagramforeachmodelwhichgivesagoodviewonthedrivingforcesandtheattractivestatesofthesystem.TheanalysisshowsthattheemergenceofminorclustersintheWeisbuch-Deffuantmodelandofmeta-stablestateswithveryslowconvergencetoconsensusintheHegselmann-Krausemodelareintrinsictothedynamicalbehavior.KEYWORDS
continuousopiniondynamics,boundedcon dence
sentableasarealnumber.Thus,agentscancompromisebyaveragingtheiropinions.Welabelsuchdynamicsascontinuousopiniondynamics.’Continuous’referstothetypeofopinionsnottothetime.Thisisincontrasttoseveralmodelsofopiniondynamicswhichcoverbinaryopinionswhereagentshavetodecide’yes’or’no’.
Examplesforcontinuousopinionsarepricesorthecon-tinuumformlefttorightinpolitics.Clustersofagentsinthecontinuousopinionspacemayrepresente.g.low-budgetvs.luxuryshoppersorpoliticalparties.
Thereareseveralmoreorlesscomplexmodelscon-cerningcontinuousopiniondynamics5,TherecentlymostdiscussedmodelsarethemodelsofWeisbuch-Deffuant(WD)andHegselmann-Krause(HK)Bothmodelsuseverysimilarheuristics.Wewillpresentbothmodelsandfocusonthedynamicaldif-ferencesincomputersimulation.
Inthethirdsectionwewillde neinteractiveMarkovChainsforbothmodels,whichleadusto’discrete’bifur-repeatedaveraging,cationdiagramsthatwillgiveusdeeperinsightsforthe
agent-basedmodels.
1Introduction2TwoAgent-basedModels
ConsideracertainnumberofagentsdiscussingacertainOpiniondynamicsintheWDandtheHKmodelisagent-issue.Eachagenthasanopinionaboutthatissueandbasedanddrivenbyrepeatedaveragingunderboundedtheytryto ndaconsensus.Theopinionshouldberepre-con dence.Themodelsdifferintheircommunication
热门文档