Epistemic semantics for counterfactuals

Abstract

Stalnaker and Lewis fixed Ramsey's suggestion into possible world semantics in effect by turning it into truth conditions. More natural, perhaps, would be to interpret it epistemically, as a belief or acceptability condition: A conditional is to be accepted in a particular body of beliefs just in case adjusting those beliefs so as to accommodate its antecedent would result in belief in its consequent. Where the conditional is a counterfactual the adjustments to be made in accommodating its antecedent would involve revision, or suspending old beliefs so as to make way for new. Relatively recently some philosophers have become interested in revision at a quite abstract level, seeking general methods for revising bodies of information and general constraints on the rationality of these methods. One such constraint is already implicit in Stalnaker's paraphrase of the Ramsey rule, where adjustments are to be made in order to maintain consistency. It would be irrational to start believing just anything or learning something new, unless of course the new belief is itself contradictory. In order to study the formal properties of belief revision, Peter Giirdenfors has introduced the notion of belief revision models, which are systems of belief states together with revision functions saying how to revise these so as to accommodate new beliefs. The Ramsey rule provides an obvious interpretation for conditionals in belief