An axiomatic approach to self-referential truth (Q1096626)

The authors investigate results of adding a `truth predicate' T( ) to Peano Arithmetic. Of course, the new axiom scheme \(A\leftrightarrow T(\#A)\) produces a contradiction. [The Gödel number {\#} will be suppressed.] But what about each implication by itself? First, a base theory is set up that includes primitive recursive arithmetic and some obvious properties of T( ) like T(A\(\to B)\&T(A)\to T(B)\). Then the authors consider eight new axioms and four rules, and make a complete classification of consistent and inconsistent combinations of them. There are only nine maximally consistent combinations. [Some samples of new axioms and rules are: \(A\to T(A)\), T(T(A))\(\to T(A)\), T(A)\(\vee T(\neg A)\), T(\(\exists xAx)\to \exists nT(An)\), and from \(\neg A\) to \(\neg T(A).]\) They provide charts to display the content of these results. Consistencies are shown by building models, some of which are constructed by stages (thus `the truth' is gradually revealed). All inconsistencies, but one, are shown by the liar paradox, essentially; the exception uses ``Schematic Löb's Theorem''. In the last section, a consistent combination is shown to be conservative over PA, and another is shown to be of the same strength as Bar Induction as to arithmetic sentences.