Believability based Iterated Belief Revision

The theory of belief revision describes how the beliefs of an agent should change upon receiving the new information. Classical iterated belief revision methods mainly focus on the consistency of belief change, with little concern of the impact of the uncertain information in multi-agent system and the process of revision. In this paper, an approach of believability based iterated belief revision is presented. This approach relates the belief revision in the multi-agent system to the believability of information, which plays an important role in the revision process. Based on the Dempster-Shafer theory of evidence and believability function formalism, the believability of information can be obtained, and thus the maximal consistent subset with the biggest believability is chosen to compose the revised belief set. The revised belief set by believability based iterated belief revision is dependent on the history of revision, namely, on the information received prior to the current belief set.