Abstract

The acceptance or rejection of computer aids to clinical reasoning is determined not only by the preferences and prejudices of potential users, but also by whether the output generated by the computer aid represents sound clinical judgment. This paper deals with the issue of the appropriate criteria for evaluating the clinical ‘reasoning’ of computer aids. Evaluation of a computer aid should include an assessment of the accuracy or appropriateness of its conclusions and an assessment of the scope of its considerations. In order to use a conclusion generated by a computer aid as a basis for clinical action, information must be available on the alternative conclusions that might have been drawn, the degree of confidence in each conclusion, and the consequences that may result from acting upon the wrong conclusion. If this information is not generated as part of the output of a computer aid, it may be obtained through the use of its explanation and display capabilities, if the aid has such capabilities. Explanation and display capabilities permit the examination of the operations of the computer aid. Examination of a computer's operations is also the most feasible means to evaluate the soundness of its ‘reasoning’