Using qualitative reasoning for diagnostic applications

Explores one way in which expert systems can be given domain knowledge to enable them to generate new diagnostic rules in the same way as human experts. This involves building a model of the domain and then comparing that model with the real world and accounting for the differences. Encouraging results have been documented for such diverse domains as electronic circuits, photocopiers and mechanical devices such as doorlocks. The last of these is used to illustrate the principles of the field. Qualitative modelling is a particular technique for building such models. The paper describes why it is both useful and limited, and explores principled ways of overcoming its limitations