Revising structured knowledge bases

In this paper we present a new approach to belief revision. In contrast to traditional formalizations of this problem, where all pieces of information included in a knowledge base have identical status, we explicitly distinguish between observations, i.e., facts which an agent observes or is being told, and facts representing a general, sometimes defeasible, knowledge about the considered world.1 This distinction allows us to deal with scenarios that cannot be properly modelled using existing belief revision operators.