Identity in modal logic theorem proving (Q687161)

In this paper the `relational' translation from quantified modal logics with identity into first-order predicate logic is repeated. In this well- known method, the Kripke semantics of the modal \(\square\)-operator: \(\square p\) is true in a world if \(p\) is true in all accessible worlds, is turned into a translation rule: \(\text{tr}(\square p,w)\to \forall v R(w,v)\Rightarrow\text{tr}(p,v)\). Predicate symbols get an extra argument which records the actual world in form of a `world term'. The method which is actually reported in the paper is somewhat more complicated because an extra predicate \(W\) is used to give the `world variables' the sort `world'. The intention behind this \(W\) predicate is to prevent the identification of world terms with domain terms. (This is in fact not necessary because the terms of the two different sorts always occur in different arguments of the predicates. In any reasonable calculus they never get mixed.) By some unmentioned reason the terms in the translated first-order modal logic version do not get extra world arguments. Therefore equations \(a= b\) can either be left as they are during the translation, and therefore they lose their modal context, or they are translated into \(I(a,b,w)\) where \(I\) is a world-dependent equality predicate. In this case the \(I\) predicate has to be axiomatized accordingly. It is not mentioned how the existential quantifier is treated in the theorem prover THINKER the author uses to prove the translated formulae, but it is very likely that both presented versions for dealing with equality cause serious problems if the existentially quantified variables occur in different modal contexts.