Validating expert systems

D.E. O´Leary (1987) noted the inappropriate nature of traditional software validation techniques for expert system validation and proposed a research methods approach. R.M. O´Keefe et al. (1987) established validation as an important concern in expert system research and development and presented general descriptive methodologies for qualitative and quantitative methods of formal validation. A specific formal paradigm for prototype validation that is consistent with O´Leary´s as well as O´Keefe´s general guidelines for expert system development is presented. The paradigm explicitly incorporates validation into the prototype development life cycle under constrained resources, primarily time and money. The validation approach is demonstrated and evaluated relative to an expert system prototype named Signal Control at Isolated Intersections. The paradigm´s strengths and weaknesses as well as its potential use with other types of expert system development projects are examined.<>