Two prototype architectures for rational interacting agents

Two of the central features that determine how an agent reasons and acts in the world are its attitudes (knowledge, beliefs, goals, values, etc.) and its architecture. In this paper, we present and discuss two alternative agent architectures than have been implemented for different purposes — one as a tool to help designers to think about systems from a user’s perspective, and the other as an argumentative agent to work in an educational setting. Both architectures embody simple principles of rationality, but the different requirements on the design of the systems resulted in the specification and implementation of subtly but substantively different architectures. We discuss some of the reasons for this and draw out some common themes for future work on the design of rational agents.