ABSTRACT
A key element of our approach is the interpretation of “self” in a meta-cognitive sense: that is, “self” is understood as a virtual character representing an agent as the subject of experience, as the target of attribution of experiences and deliberate actions performed by this agent. Thus understood, “self” can be represented as an element in an agent’s cognitive system and can be used for meta-cognitive processing: i.e., reasoning about one’s own self and other selves. This general idea reflects a simulationist theory-of-mind viewpoint (Nichols & Stich, 2000), which is taken as the basis for our approach. Our model of an agent’s mind includes multiple instances of “self” representing notions of I-Now, I-Yesterday, I-Imagine, I-Goal, etc. Each instance of “self” is represented by a “chart” with a set of properties and mental states attributed to it. Thus, mental states in this framework are representations of experiences attributed to a particular instance of “self”. This attribution further implies certain rules and constraints imposed on the contents and the dynamics of representations. The result is a general architecture that will enable in intelligent agents a meta-cognitive “common sense”, which proves to be vital in a variety of paradigms and scenarios requiring cooperation within a team.
