Modeling medical diagnosis: Logical and computer approaches

Conclusion In the present article I have surveyed several approaches to modeling the clinical diagnostic process. I have argued that at this point of the field's development, logics which simulate the reasoning patterns and knowledge base of expert clinicians represent research programs that are most likely to succeed. No logic of diagnosis has yet attained the status of being definitive; in spite of striking progress much more research and testing is required. On the basis of various existing logics, I have attempted to articulate a number of desiderata which an ideal diagnostic logic should satisfy. In spite of these criticisms, I would maintain that programs such as MYCIN and INTERNIST have already matured to the point where they are useful both in consultations and in teaching diagnostic reasoning.Department of History and Philosophy of Science, University of Pittsburgh, Pittsburgh, PA 15260. The author wishes to acknowledge gratefully the assistance of Drs. Jack D. Myers, Harry E. Pople, Jr., Joshua Nagin, and Randy Miller with the research that resulted in this essay and thanks to L. J. Cohen, J. F. Fries, R. N. Giere, and C. Whitbeck for comments on an earlier draft. Supported in part by training grant MB 00199-01 from the Bureau of Health Manpower (DHEW) and by the National Science Foundation and the National Endowment for the Humanities.Presented at a symposium on Philosophy and Medicine held at the American Philosophical Association's Eastern Division Meeting in Washington, D.C., December 27–29, 1978. This essay partially overlaps with an article on Problems in computer diagnosis to appear inThe Ethics of Diagnosis, H. T. Engelhardt, Jr. and S. Spicker (eds.) Dordrecht: D. Reidel, forthcoming.