概率公共信念的动态刻画.pdfVIP

StudiesinLogic,Vol.8, No. 4 (2015): 32–53 PII:1674-3202(2015)-04-0032-22 * DynamicsforProbabilistic Common Belief EricPacuit DepartmentofPhilosophy,University of Maryland epacuit@ Abstract. Robert Aumann’s agreeing to disagree theorem shows that if two agents have the same prior probability and update their probability of an event with private information by conditioning,theniftheposteriorprobabilitiesof arecommonknowledge,thentheposteriors must be the same. Dov Monderer and Dov Samet prove a generalization of Aumann’s result involving a probabilistic variant of common knowledge. In this paper, I use various methods from probabilistic and dynamic-epistemic logics to explore a dynamic characterization of the Monderer and Samet result. The main goal is to develop a model that describes the evolution of the agents’ knowledge and (probabilistic) beliefs as they interact with each other and the environment. I will show how the logical frameworks are rich and flexible tools that can be usedto study many dynamic processesofinteractivesocialbeliefchange. 1 Introduction In 1976, Robert Aumann proved a fascinating result ([3]). Suppose that two agents have the same prior probability and update their probability of an event withprivateinformationbyconditioning. Aumannshowedthatiftheposteriorprob- abilitiesof arecommonknowledge,thentheagentsmustassignthe same posterior to . This is true even if the agents receive different information. In other words, ifagentshavethesamepriorprobabilityandupdatebyconditioning,thentheagents cannot“agreetodisagree”abouttheirposteriorprobabilities. Thisseminalresulthas beengeneralizedinmanyways([4, 12,41,42])andisstillthesubjectofmuchdiscus- sion in Economics (