Knowledge-level assessment of problem solving methods

The strong interest in knowledge level description of expert systems has led to various efforts defining and cataloguing AI problem solving methods and building methodologies and tools to use them. We address the issue of matching problem solving methods to tasks. This includes: deciding whether a method is applicable to the problem at hand; if it is applicable, assessing the extent of applicability in terms of quality of results, reliability, and so forth; and deciding how to configure the method for the problem at hand. The three issues are addressed in turn. We develop a documentation framework identifying all the features of the problem solving method and the features of the problem to which it is applied. We develop a procedure for analyzing these features and deriving an assessment of the scope and the quality of the application. We show how the analysis process produces a set of problem configuration guidelines helping system designers find the optimal way to model the problem.