For automatic theorem provers it is as important to disprove false conjectures as it is to prove true ones, especially if it is not known ahead of time if a formula is derivable inside a particular inference system. Situations of this kind occur frequently in inductive theorem proving systems where failure is a common mode of operation. This paper describes an abstraction mechanism for first-order logic over an arbitrary but fixed term algebra to second-order monadic logic with 0 successor functions. The decidability of second-order monadic logic together with our notion of abstraction yields an elegant criterion that characterizes a subclass of unprovable conjectures.
展开▼