Analogical reasoning for second generation expert systems

Abstract

In this paper we give a critical view on current rule oriented expert systems and show that there are arguments against the rule oriented strategy from the practical as well as the theoretical point of view. Instead of this strategy we propose a formalization of analogical reasoning, which leads for second order expert systems to a case oriented strategy. In some areas of artificial intelligence certain approaches to the formalization of analogy exist. But a unified mathematical theory of analogy is needed. After a short description of a possible general mathematical approach to analogical reasoning we will describe some applications and implications of this new method. The potential applications for analogical reasoning have one difficulty in common: the used knowledge is to represent in a structural way for detecting similarities and analogies on a certain level.