Abstract
Mindreading is the ability to understand both oneself and other agents in terms of beliefs, desires, intentions and other relevant mental states. This critically important ability has been implicated across a broad spectrum of human cognitive activities, including participation in dialogue, collaboration, competition, and moral judgment. This paper serves as a reflection on the kinds of strategies we can use to build a system capable of mindreading, given currently available resources in the relevant literature. After reviewing representative computational approaches on offer I will suggest a set of architectural mechanisms that could provide the flexibility required to build a robust mindreading capability for cognitive systems.
