A method for validating expert systems, based on psychological validation literature and Turing's "imitation game," is applied to a flexible benefits expert system. Expert system validation entails determining if a difference exists between expert and novice decisions (construct validity), if the system uses the same inputs and processes to make its decisions as experts (content validity), and if the system produces the same results as experts (criterionrelated validity). If these criteria are satisfied, then the system is indistinguishable from experts for its domain and satisfies Turing's "imitation game."
The methods developed in this paper are applied to a human resource expert system, Personal Choice Expert (PCE), designed to help employees choose a benefits package in a flexible benefits system. Expert and novice recommendations are compared to those generated by PCE. PCE's recommendations do not significantly differ from those given by experts. High inter-expert agreement exists for some benefit recommendations (e.g. Dental Care and Long-Term Disability) but not for others (e.g. Short-Term Disability and Life Insurance). Insights offered by this method are illustrated and examined.