Architectures and Tools for Human-Like Agents

This paper discusses agent architectures which are describable in terms of the “higher level” mental concepts applicable to human beings, e.g. “believes”, “desires”, “intends” and “feels”. We conjecture that such concepts are grounded in a type of information processing architecture, and not simply in observable behaviour nor in Newell’s knowledge-level concepts, nor Dennett’s “intentional stance.” A strategy for conceptual exploration of architectures in design-space and nichespace is outlined, including an analysis of design tradeoffs. The SIM AGENT toolkit, developed to support such exploration, including hybrid architectures, is described briefly.