Counterexamples to recovery and the filtering condition (Q1871129)

The author participates in an ongoing discussion of the status of the postulate of recovery, as a condition on contraction from a belief set closed under classical consequence. On the one hand, the AGM constructions satisfy it, but on the other hand there are examples that suggest that it is inappropriate. In an earlier paper, the reviewer had suggested that we should distinguish between belief states understood as bare sets of propositions closed under consequence, and such sets equipped with additional justificational structure. Recovery, he argued, is appropriate for the former but not the latter. In the paper under review, the author criticises that position, contending that even in the former context recovery is not acceptable. The subset relation, he contends, is a poor guide to informational value, since two sets of propositions, both closed under consequence, may have the same informational value even though both are closed under consequence and one is a proper subset of the other. The AGM definition of partial meet contraction may therefore profitably be re-run, he suggests, by replacing the subset relation by a suitably defined or constrained relation of relative informational value. As the author has pointed out in previous publications, this will in general fail recovery.