Relevant implication and the case for a weaker logic (Q1914376)

This paper starts by rehearsing a number of problems with the system of relevant implication, R. It argues, instead, for a weaker relevant system, called DJ\(^d\). The majority of the paper provides a semantics for this system and proves soundness and completeness. The central notion of the semantics is that of the content of a sentence \(\alpha\), \(c(\alpha)\), which is, intuitively, the set of sentences that \(\alpha\) entails; and a conditional \(\alpha\to\beta\) is true in an interpretation iff \(c(\alpha)\supseteq c(\beta)\). It is doubtful that this provides an independent justification for the logic, however, since various conditions governing the behaviour of contents are postulated, and many postulates are little more than the axioms of the logic in semantic form. The final part of the paper discusses how classical logic can be recaptured in DJ\(^d\). A separate style of syntactic variable is introduced, and it is required semantically that these satisfy the laws of excluded middle and non-contradiction; the rest of classical logic for them then follows. This approach gives a certain problem with applying the logic, however, since in a natural language, there is no syntactic marker of classicality in this sense. For example, the paper suggests that the sentences of the metatheory given are classical; but since the language is that of set theory, and one of the prime purposes of the logic is to provide for a non-classical set-theory, it is not at all clear why this should be so.