Communication Breakdown? Reasoning about Language and Rational Agents

This paper examines different approaches to rationality in analytic philosophy and AI, in the light of Bermúdez’s proposal that a full account of rationality must aim to explain how agents can both select and explain actions, as well as assessing them against some normative standard. We briefly survey instrumental, linguistic and discursive accounts of rationality, and conjecture that Habermas’s notion of the “three roots” of epistemic, teleological and communicative rationality comes closest to providing a satisfactory account, or at least the ingredients of such an account. This is contrasted with the widely-accepted BDI model of rational agency in AI, which we argue falls short of a full model of rationality and in particular, fails to provide a convincing model of linguistic communication.