Salem Benferhat, Sébastien Konieczny, Odile Papini, Ramón Pino Pérez
Epistemic states are suitable to represent intelligent agent's knowledge. This paper discusses iterated revision of epistemic states by epistemic states from axiomatic, semantics and syntactic points of view. We first propose a set of general postulates that such iterated revision operations should verify, independently from the epistemic states representation. We give a representation theorem for a class of revision operation by epistemic states called "minimally-preserving models" revision operators. Then, at the semantical level, we propose to use polynomials for encoding epistemic states. We show that this encoding makes easy the iteration of the revision process using simple operations on these polynomials. Lastly, we give an efficient implementation of revision by epistemic states, when the later are syntactically represented by prioritized knowledge bases. The revision by epistemic states proposed in this paper generalizes (axiomatically, semantically and syntactically) iterated revision operations by (a single and non-prioritized) formula, stemming from the principle of strong primacy of new information.
Keywords: Belief Revision, Knowledge Representation, Nonmonotonic Reasoning
Citation: Salem Benferhat, Sébastien Konieczny, Odile Papini, Ramón Pino Pérez: Iterated Revision by Epistemic States: Axioms, Semantics and Syntax. In W.Horn (ed.): ECAI2000, Proceedings of the 14th European Conference on Artificial Intelligence, IOS Press, Amsterdam, 2000, pp.13-17.