Belief Generalization for Intelligent Agents

Abstract

One of the important characteristics for intelligent agents is to be able to assess their environments in order to generate correct beliefs to make right decisions. It is always difficult to do so because many factors including uncertain information, knowledge and bounded time will affect intelligent agents to perceive their environments. In this paper, we propose a procedure descriptive framework rather than a logical model to describe how agents automatically form their beliefs. The process of belief updating in this framework remains to be constantly changing, until the point of decision making is reached. Using this framework, a subject agent can smoothly fuse its knowledge and possible information from the current environment in order to obtain correct beliefs. We also present a solution for conflict resolution between agents in this framework.