Abstract

This dissertation discusses rational strategies for eliminating from a body of knowledge any inconsistency introduced by new evidence. It is assumed that an agent judiciously gives up some of her beliefs before accepting any contravening evidence. The guiding principle behind this discussion is that information being costly, its loss should be minimized during this purgative process. The research work is done within the AGM framework of belief revision developed by Peter Gardenfors and his colleagues. ;The first chapter introduces the salient concepts that are used later in the dissertation. The second chapter is concerned with the construction of a base contraction operation such that the theory contraction operation generated by it will be fully AGM-rational. It is shown that the theory contraction operation generated by Fuhrmann's minimal base contraction operation, even under quite strong restrictions, fails to satisfy the "supplementary postulates" of belief contraction. Finally Fuhrmann's construction is appropriately modified so as to yield the desired properties. The new construction may be described as involving a modification of safe contraction so as to make it maxichoice. ;The next chapter is devoted to a philosophical defense of the principle of recovery. One of the recurrent themes of this chapter is that the AGM representation of the new information is too simplistic, and that its proper representation should take into account the context in which this information is acquired. It is further argued that the arguments normally marshaled against recovery fail when information is thus represented. ;In the final chapter, a belief revision operation is constructed that, unlike the AGM revision operations, is capable of handling multiple belief revision and iterated belief revision. For this purpose, both the body of knowledge and the new information are represented as epistemic entrenchment relations. This revision operation may roughly be seen as a generalization of the AGM revision operations, as Jeffrey conditionalization is a generalization of the orthodox Bayesian conditionalization.