A Knowledge Level Analysis of Belief

Revising beliefs is a task any intelligent agent has to perform. For this reason, belief revision has received much interest in Artii-cial Intelligence. However, there are serious problems when trying to analyze belief revision techniques developed in the eld of Arti-cial Intelligence on the knowledge level. The symbolic representation of beliefs seems to be crucial. The theory of epistemic change shows that a partial knowledge-level analysis of belief revision is possible, but leaves open the question of how this theory is related to belief revision approaches in Artii-cial Intelligence. In particular, it remains an open question whether the results achieved in the knowledge-level analysis are valid. Furthermore , the idea of reason maintenance, which is considered to be essential in AI, has no counterpart in the theory of epis-temic change. Addressing these problems, it is shown how to reconstruct symbol-level belief revision on the knowledge level.