An Agent Memory Model Enabling Rational and Biased Reasoning

This paper presents an architecture for a memory model that facilitates versatile reasoning mechanisms over the beliefs stored in an agentpsilas belief base. Based on an approach for belief aggregation, a model is introduced for controlling both the formation of abstract and complex beliefs and the retrieval of them based on their activation history. An implementation of the presented mechanisms illustrates how it can be used in intelligent agents that exhibit human-like (biased) reasoning as well as rational reasoning.