A priori revision

Abstract

The problem of revision is to find which formula ψ can be deduced from a formula φ, which has been added to a Knowledge Base KB. Since φ can bring inconsistency to KB, non-monotonic inference relations which are able to deal with inconsistency have been proposed; note that classical revision takes place after the arrival of φ. The aim of this paper is to propose a priori revision, that is to provide a way to "armor" the KB by suppressing some knowledge and by forbidding to accept some new information in such a way that adding any allowed formula φ to the revised KB will not bring inconsistency. © Springer-Verlag Berlin Heidelberg 2001.