Hypothesis generation by machine

摘要

A method for deriving sets of falsehood preserving inference rules (for first-order predicate calculus) from truth preserving inference rules is demonstrated. It is shown that such sets of rules have a property of completeness analogous to the completeness of the truth preserving rules. Further, it is demonstrated that falsehood preserving rules lead from consequences to axioms, and thus they can be used for hypothesis generation. In particular, a deductive consequence generating machine may be used as an hypothesis generating machine if one employs the methods outlined.