Belief Ascription and Mental-level Modelling

Models of agents that employ formal notions of mental states are useful and often easier to construct than models at the symbol (e.g., programming language) or physical (e.g., mechanical) level. However, to enjoy these benefits, we must first supply a coherent picture of mental-level models. What is required is a description of the various components of the mental level, their dynamics, their interrelations, and their relations with the agent's behavior. Only then will we have a complete semantics for mental notions. The goal of the first part of this paper is to provide this picture. The second part of this paper concentrates specifically on belief ascription. We address two fundamental unresolved problems. Our mental-level model addresses the question of grounding: where do beliefs come from, i.e., what links a system's symbol or physical level with its beliefs? Our characterization of a class of goal-seeking agents goes towards addressing the question of adequacy: when can we treat an entity as having beliefs? In addition, we look at general assumptions that can help constrain the set of beliefs an agent can be ascribed. Together, these results supply a basis for agent modelling using mental states.