No abstract

Abstract. Generic models have received widespread attention in knowledge based systems research (KBS) as an important aid in the process of modeling "problem solving behavior". However, little empirical evidence has been presented which justifies this role. In this article, we analyze some of our experiences with using generic models of problem solving in industrial projects. In these projects, we have relied upon generic models to support the acquisition and modeling process as advocated by current modeling methodologies. Based on these experiences, we argue that there is a real danger that practitioners might be misleaded by inappropriate generic models because current modeling methodologies do not sufficiently address the issues related to selecting, refining and instantiating generic models. To avoid this costly endeavor, we argue that modeling methodologies should emphasize more the nature of the domain, the work that people do, the role of the environment and an active role of the expert. We also emphasize the importance of a tight coupling between the modeling activity and validation. Validation should be carried out with respect to the expert source and the contribution of each model construct towards the intended goal of the models. The paper concludes with remarks on a framework that we call "validation through remodeling" .