18 - Non-Monotonic Reasoning in Interpretation (Update of Chapter 17)

Abstract

This chapter presents recent progress made in formal pragmatics by using nonmonotonic logic to describe pragmatic meaning. It highlights how minimal models can be used to describe and explain inferences of language use, in particular Gricean conversational implicatures. The chapter discusses how nonmonotonic logic can be used at the semantic–pragmatic interface to account for the preferred interpretation of a sentence. The last part of the chapter discusses the role of non-monotonic logic for human reasoning in general. When McCarthy discusses possible applications of nonmonotonic logic in general, and circumscription in particular, he also mentions communicational conventions. The example he provides actually involves Grice's maxims of conversation. Given this observation, the recent developments in the area appear to mark only a shift in perspective. In the 1970s and particularly the 1980s in AI one came across some interesting questions concerning reasoning and intelligent behavior in general. This led to the development of very sophisticated logical tools to answer these questions. At the same time similar problems have also been discussed in linguistics and psychology, but without the means of developing a formal tool to deal with them. At a certain point in the 1990s the interests of AI in nonmonotonic logic weakened, because the conviction emerged that nonmonotonic logic is not fit to deal with modeling the relevant phenomena in an efficient way. But now linguistics and psychology are taking over, with a more sophisticated understanding of communication. Nonmonotonic logic is rediscovered as a very useful tool to describe and explain certain aspects of interpretation and reasoning.