Assessing the Minimality of Change in Belief Revision: Capacities

Abstract

The process of belief revision [Alchourrón et al., 1985] was developed to model the effect of accepting incoming information into a knowledge base. It can be based on a total preordering of the set of beliefs held by the system. When this ordering relates in a natural way to the semantics of the underlying knowledge base, it is dubbed an epistemic entrenchment ordering [Gärdenfors and Makinson, 19881 This ranks beliefs in order of the agent’s reluctance to give them up. Thus the belief the sun will rise tomorrow would be ranked higher than the universe started with a big bang. Such an ordering leads to the construction of a unique AGM transformation on the knowledge base to allow for the acceptance of new information; information which may be inconsistent with the current beliefs. This original work was purely theoretical, concerning itself only with one revision. The outcome was a revised set of beliefs, but the question of how those beliefs could be fitted with a “revised” ranking was not addressed. This question is, however, not only theoretically interesting but also vital for implementation since, in that situation, it is necessary to iterate the procedure and so a new epistemic entrenchment needs to be the outcome of each revision.KeywordsBelief RevisionExtension CapacityEpistemic EntrenchmentBase RevisionIterate Belief RevisionThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.