Where there’s a Will there’s an agent

Abstract

This theoretical paper introduces a formal architecture that supports the integration of various A.I. technologies, including planning, to make an autonomous agent. The design decisions made are self-consciously minimal, resulting in a simple but general architecture. Its functionality appears comparable to other agent architectures in the literature, including the incorporation of a reactive element. Novelties in the architecture include “cognitive reactivity,” a formal notion of relevance, and attention, and a control regime based on these. An initial implementation is briefly described.