MODELING BELIEF CHANGE IN A POPULATION USING EXPLANATORY COHERENCE

Abstract

A simulation model that represents belief change within a population of agents who are connected by a social network is presented based on Thagard's theory of explanatory coherence. In this model there are a fixed number of represented beliefs, each of which are either held or not by each agent. These are conceived of existing against a background of a large set of (unrepresented) shared beliefs. These beliefs are to different extents coherent with each other — this is modeled using a coherence function from possible sets of core beliefs to [-1, 1]. The social influence is achieved through gaining of a belief across a social link. Beliefs can be lost by being dropped from an agent's store. Both of these processes happen with a probability related to the change in coherence that would result in an agent's belief store. A resulting measured "opinion" can be retrieved in a number of ways, here as a weighted sum of a pattern of the core beliefs — opinion is thus an outcome and not directly processed by agents. This model suggests hypotheses about group opinion dynamics that differ from that of many established models.