Equilibrium Linguistics: Computing content

Prashant Parikh, University of Pennsylvania and Noema

Abstract

I will describe a framework that enables one to derive the content of an utterance from first principles by modeling it as a system of interdependent games. I will suggest that this way of computing meaning allows us to unify semantics and pragmatics into a single discipline in which literal meaning and implicature co-determine each other. I include syntactic (and phonological) content in this computation and will show how treating (phonology and) syntax and semantics simultaneously leads to an interesting and more general way of thinking about compositionality. The key reason why game theory is so widely applicable is that many aspects of communication can be seen as problems involving disambiguation, and game theory is really about agents making interdependent choices among alternatives. It may be possible to extend this picture to all of language implying that it involves many processes that are circular and reciprocal rather than linear and sequential. If this turns out to be right, equilibrium may be as central an idea for the science of language as generativity.