Iterated Belief Revision: A Computational Approach

The capability of revising its beliefs upon new information in a rational and efficient way is crucial for an intelligent agent. The classic AGM theory studies mathematically idealized models of belief revision in two aspects: the properties (i.e., the AGM postulates) a rational revision operator should satisfy; and how to construct concrete revision operators. In scenarios where new information arrives in sequence, rational revision operators should also respect postulates for iterated revision (e.g., the DP postulates). When applications are concerned, the idealization of the AGM theory has to be lifted, in particular, beliefs of an agent should be represented by a finite belief base. In this paper, we present a computational base revision operator, which satisfies the AGM postulates and postulates for iterated revision. We will show that our base revision operator is almost optimal in terms of computational complexity. Furthermore, the base revision operator’s degrees of syntax relevance and minimal change are also formally analyzed. Topic Area: Iterated Base Revision, Computational Complexity, Syntax Irrelevance, Minimal Change Acknowledgments The first author is funded by the Deutsche Forschungsgemeinschaft under grant no. Gr 334/3. Dagstuhl Seminar Proceedings 05321 Belief Change in Rational Agents: Perspectives from Artificial Intelligence, Philosophy, and Economics http://drops.dagstuhl.de/opus/volltexte/2005/359