Proof-complexity results for nonmonotonic reasoning

Abstract

It is well-known that almost all nonmonotonic formalisms have a higher worst-case complexity than classical reasoning. In some sense, this observation denies one of the original motivations of nonmonotonic systems, which was the expectation taht nonmonotonic rules should help tospeed-upthe reasoning process, and not make it more difficult. In this paper, we look at this issue from a proof-theoretical perspective. We consider analytic calculi for certain nonmonotonic logis and analyze to what extent the presence of nonmonotonic rules can simplify the search for proofs. In particular, we show that there are classes of first-order formulae which have only extremely long “classical” proofs, i.e., proofs without applications of nonmonotonic rules, but there are short proofs using nonmonotonic inferences. Hence,despite the increase of complexity in the worst case, there are instances where nonmonotonic reasoning can be much simpler than classical (cut-free) reasoning.