Agents that Rationalize their Decisions

This paper presents an intr{gluction to a theory of rationalization for autonomous agents covering three kinds of rationalization: goal-bas~xl, belief based, and assumption-based rationalization. Wc show that the COMINT model of decisioa making (Botclho & (’oelho 1996) enables us to approach rationalization in a rather natural fashion. Namely, it presents an automatic (as opposed to deliberative and thoughtful) way of selecting the convenient sets of knowledgc structures to rationalize the agent’s decisions. This automatic mechanism relies on the concepts of activation and association, both of which are central to the model of memory (Botelho & (’oelho 1995) underlying COMINT. We also give a fiwmal definition of a rationalization basis anti show how it is generated. Along the paper, wc discuss some examples of personnel selcction in organizations. Some aspects of the implementation of the model are also presented. Although rationalization serves a lot of purposes, this paper emphasizes its role in a multi-agent society.