What is a relevant connective? (Q2155944)

The centerpiece of the family of \textit{relevant logics} (see [\textit{A. R. Anderson} et al., Entailment. The logic of relevance and necessity. Vol. I. Princeton, NJ: Princeton University Press (1975; Zbl 0323.02030)] for a survey) is the binary connective \(\rightarrow\), so that \(\varphi\rightarrow\psi\) is interpreted as indicating that a relation of \textit{entailment} holds between statements \(\varphi\) and \(\psi\). While a number of formal constraints purporting to characterize the eponymous `relevance' of the connective have been offered, it is generally accepted that satisfying the \textit{variable-sharing property} (that theoremhood of \(\varphi\rightarrow\psi\) requires that some atomic formula appears in both \(\varphi\) and \(\psi\)) is a necessary condition on such a logic. Despite the emphasis on the connective \(\rightarrow\), one can ask about the degree to which other sentential connectives cohere (or not) with the underlying motivations for relevant logics. For example, that the introduction of a new connective \(\star\) to intuitionistic logic should cause the failure of its characteristic \textit{disjunction property} (that \(\varphi\vee\psi\) is provable only if \(\varphi\) or \(\psi\) is provable) might be taken as evidence that \(\star\) is not appropriate as an `intuitionist' connective. The author takes up this question in the case of relevant logics by investigating the formal conditions that a connective must satisfy if it is to be consistent with their underlying philosophical ideology. The author takes as a starting point the five-point characterization of `intuitionistic connectives' drawn from [\textit{D. M. Gabbay}, Semantical investigations in Heyting's intuitionistic logic. Dordrecht; D. Reidel Publishing Company (1981; Zbl 0453.03001)], discussing the applicability of particular clauses to the relevant case. After surveying the philosophical literature on relevant logics and the ternary relation semantics described in [\textit{R. Routley} and \textit{R. K. Meyer}, Truth, Syntax, Modality, Proc. Temple Univ. Conf. Alternat. Semantics 1970, Studies Logic Foundations Math. 68, 199--243. (1974; Zbl 0317.02017)], the author offers three criteria providing a preliminary account of a connective \(\star\)'s being `relevant.' These criteria require that the introduction of \(\star\) to a relevant logic does not: \begin{itemize} \item cause the failure of the variable-sharing property \item lead to theoremhood of `weakening-like' formulae \(C(\varphi)\rightarrow (\psi\rightarrow D(\varphi)\) where \(C(\_)\) and \(D(\_)\) are schemata such that neither \(\varphi\), \(C(\varphi)\), nor \(C(\varphi)\) share variables with \(\psi\) \item violate heredity in the ternary relation model theory (\textit{i.e.} if \(s\Vdash\varphi\) and \(s\leq t\) then \(t\Vdash\varphi\)) \end{itemize} The author proceeds to investigate model-theoretic presentations of a host of plausible connectives. Importantly, the author notes that connectives that lead to \textit{ubiquitous formulas} -- those that will either be verified at every point in a model or falsified at every point in a model -- must be rejected insofar as such ubiquity will immediately lead to violations of variable-sharing. The author notes that connectives in which the point of evaluation plays no role in their truth conditions, \textit{i.e.}, that can be given conditions \[ s\Vdash \star(\xi_{0},..f.,\xi_{j})\text{ if }\Phi,\text{ where }s\text{ is not mentioned in }\Phi \] will lead to ubiquitous formulas. Likewise, the author observes that connectives whose truth conditions can be described in terms of set-theoretic relations between the sets of points verifying its subformulas will lead to variable-sharing as well. \textit{I.e.}, if \([\psi]\) is the set of points in a model verifying \([\psi]\), this rules out connectives for which a condition of the form \[ s\Vdash \star(\xi_{0},..f.,\xi_{j})\text{ if }\Phi,\text{ where }\Phi\text{ is a set-theoretic condition on sets }[\xi_{0}],...,[\xi_{j}] \] can be given. These provide two broad and elegant characterizations of connectives whose model theoretic interpretations decisively reveal them to be inconsistent with relevant intuitions. To these, the author offers a third informal heuristic, noting that if heredity of a formula \(\star(\xi_{0},...,\xi_{j})\) can be proven inductively without making use of a basis step, the connective \(\star\) will fail to meet one of the criteria. The author concludes by examining the \textit{Ackermann constant} \(\mathbf{t}\) and \textit{Church constant} \(\top\), two sentential constants often encountered as technical devices in the relevant logic literature. Given the provability of \textit{e.g.} \(\mathbf{t}\rightarrow (p\rightarrow p)\) in \(\mathbf{R}\) -- despite no common variable in \(\mathbf{t}\) and \(p\) -- variable-sharing is challenged and retained by an informal interpretation of \(\mathbf{t}\) as the infinitary conjunction of `all theorems' (and \(\bot\) as the conjunction of `all formulas'). By noting that \(\mathbf{t}\) and \(\top\) differ with respect to whether their introduction licenses ubiquitous formulas, the author's discussion allows an analysis providing reasons to consider \(\mathbf{t}\) to be `relevant' but to reject \(\top\) as conflicting with relevant motivations.