Should we be surprised at expert system success ?
Authors: Powell, P.
Journal: Information Systems Journal (1992)
DOI: 10.1111/j.1365-2575.1992.tb00082.x
<jats:p><jats:bold>Abstract. </jats:bold> Expert systems proliferate. They are now a usable and used tool in a variety of organizations. However, little research has been undertaken on the quantification of success or failure of expert system projects. By and large the only test carried out for expert system validation is one of comparison of the conclusions of the system with those of an expert or team of experts. While questioning the validity, this paper accepts that this method is being used and analyses whether or not the results obtained from this type of test should be a surprise. That is, what sort of results should be expected and to what extent do the reported ones live up to these expectations? Earlier research from the field of human information processing casts doubts on the success claims made for current expert systems. The implications of this research are then analysed in order to explore the possible redeeming features of expert systems which are not taken into account in such comparisons. The use of linear models as comparators for expert systems may prove beneficial.</jats:p>