Methodological principles in medical knowledge programming: part I

摘要

The impact of methodology on knowledge programming in medicine has been, until now, rather scarce. The reasoning strategies of most systems are variations on probabilistic themes like the BAYES formula. An analysis of the INTERNIST-I/CADUCEUS project of Myers and Pople shows that there are serious shortcomings in the probabilistic approach that impair the knowledge representation as well as the correctness of the clinical conclusions. An outline is given of how more adequate representations and decisions may be arrived at on the basis of logico-methodological reconstructions of clinical knowledge and reasoning.