An empirical model for the evaluation and selection of expert system shells

Abstract Expert systems (ES) represent a relatively new approach to solving complex problems with computers. Since the early 1980s, the introduction of ES shells has been a major factor in making the ES technology an affordable application of artificial intelligencefor most organizations. Currently, there are more than one hundred commercial ES shells available in the market. The selection of a specific ES shell for a specific application is an important decision. If the “wrong” ES shell is selected it could result in an inefficient, or ineffective, expert system or even in project failure. The availability of a large number of rapidly evolving ES shells on the market and the lack of industry standards or benchmarks, along with the lack of user experience with this technology, make the comparison and selection of an appropriate shell a difficult task. In this paper we describe an ES shell evaluation model along with guidelines for the evaluation process. This model is based on the experiences of 271 knowledge engineers and end-users. Given a particular application type, the model can be used to identify critical ES shell attributes and capabilities, which can then be used as evaluation criteria against the pool of candidate ES shells.