Propositional relevance through letter-sharing (Q846520)

The author discusses, in this paper, different notions of relevance between classical propositional formulae and examines the application of some of these notions within contexts of belief change. The first notion he takes into account is syntactic relevance, that is, relevance defined as letter-sharing. He notices that this sort of relevance is syntax-dependent. To show how this and other difficulties can be overcome, the author introduces other notions of relevance. He first considers one based on the idea that there is a unique least set of elementary letters for every set \(A\) of formulae, such that \(A\) may equivalently be expressed using only letters from that set. He calls this notion ``essential relevance''. In contrast to syntactic relevance, essential relevance is syntax-independent. When belief change is taken into account, the author points out that relevance or irrelevance of one formula to another may be taken to depend not only on the formulae themselves but also on the choice of a background belief set. This intuition leads to the notion of path relevance, which (as the author notes) generalizes essential relevance. Path relevance is syntactically independent in connection with the letters involved, but it is syntax-dependent with respect to the background belief set assumed by this notion of relevance. If we apply the notion of finest splitting of a belief set, this dependence can be overcome. Application of this notion results into what the author calls ``canonical relevance'' (modulo a belief set). This sort of relevance is syntax-independent both with respect to letters and a background belief set. However, as the author himself proves in this paper, canonical relevance is not fully language-independent, that is, it is not invariant under different representations of the same state of affairs (even when the representations are in the same language). Certain interesting interactions of canonical relevance with operations of belief change are also shown. We should point out that the aforementioned notions (of syntactic, essential, path and canonical relevance) were not originally proposed by Makinson, as he himself notes. In the final part of the paper, the author presents a new definition of relevance of his own, viz.: parameter-sensitive relevance. This notion offers a response to some of the issues indicated by the author regarding the interaction of canonical relevance with operations of belief change.