Generalized Problem-Solving Methods

Problem-Solving Methods (PSMs) describe the dynamic reasoning behaviour of a knowledge-based system independent from a certain domain and application. Evidently, the reuse of PSMs across different domains or applications is a challenging issue. The investigation presented in this paper was directed by the hypotheses that a library of PSMs should be essentially more than just a juxtaposition of its items, and that the fundamental principle for reuse is abstraction. To meet these goals the notion of a Generalized Problem-Solving Method (GPSM) is introduced. A GPSM is a model that collects information about similar PSMs, a family of PSMs. At the same time it describes the members of the family and their relations. The approach suggests these abstractions from several PSMs as a more appropriate candidate for reuse retrieval. A GPSM contains some degrees of freedom, interpretable as parameters, but it allows also to model separately how these are resolved. So, after the selection of a GPSM this approach supports a systematic way how to derive a member of the family. To this extent, the reuse of PSMs might already be seen as an engineering discipline. The main aspects of the approach are incrementally developed in this paper and always related to a case study for the family of assignment (rsp. parametric design). This is finally presented completely.