Formal description of objects, processes, and levels of expert reasoning

Abstract

In this paper an expert system is described which supports the medical diagnosis of ear-diseases. Its purpose is not only to supply the final diagnosis but also to guide the user through the whole process of examinations and tests which lead to the final decision. The situation described here is simple under the aspect that the knowledge base is not too big and can be relatively clearly structured. The authors nevertheless feel that the methods and concepts presented are of some general interest and of use in other applications. The situation we analyze is mainly described as follows:

• 1. Case related knowledge is often inherently incomplete and insufficient for the intended decisions

• 2. Domain related knowledge is often vague

• 3. New knowledge about a case may be (and remain) inconsistent with previous knowledge

• 4. Reasoning may lead to conclusions which are only valid in a sense other than formal logic

• 5. Reasoning is not only based on the present confirmed positive knowledge but may involve the absence of contradictory evidence; hence it may be non-monotonic

• 6. Knowledge has to be structured in order to allow adequate use in different contexts

• 7. Most of the meta-knowledge about the given situation is not formalized and difficult to formalize