How to combine data abstraction and model refinement: A methodological contribution in MACAO

This paper deals with methodological aspects of knowledge acquisition and modelling. We focus on how the problem solving can be modelled. Our analysis relies on two experiments where we combined MACAO and KADS to develop knowledge based systems: a technical diagnosis support application and a system that helps to assess debt recovery files. The paper reports these experiments as well as the conclusions drawn. Their evaluation underlines the advantage of combining a detailed analysis of the expert's reasoning with the selection and adaptation of generic models and problem solving methods. Moreover, from this work, we derive guidelines on how to apply practically this combination. We propose to integrate these results in MACAO and improve the methodology by this means.