A computational theory of belief introspection

Abstract

Introspection is a general term covering the ability of an agent to reflect upon the workings of his own cognitive functions. In this paper we will be concerned with developing an explanatory theory of a particular type of introspection: a robot agent's knowledge of his own beliefs. The development is both descriptive, in the sense of being able to capture introspective behavior as it exist; and prescriptive, in yielding an effective means of adding introspective reasoning abilities to robot agents.