Psychiatric Diagnosis from the Viewpoint of Computational Logic

While medical information systems have become common in the United States present systems have mostly addressed clerical aspects of medicine such as billing, record managementand similar tasks. Deeper problems, such as aiding the process of diagnosis, have largely remmained unexplored for commercial systems. This is not surprising since automating diagnosisrequires considerable sophistication both in the understanding of psychiatric epidemeology andin knowledge representation techniques. This paper is an interdisciplinary study of how recent results in logic programming, non-monotonic reasoning, and knowledge representation can aidin psychiatric diagnosis. We argue that to logically represent psychiatric diagnosis as codified in the Diagnostic and Statistical Manual of Mental Disorders, 4th edition requires abduction over programs that include both explicit and non-stratified default negation, as well as dynamic preference rules. We show how such programs can be translated into abductive frameworks over normal logic programs and implemented using recently introduced logic programming techhniques. Finally, we describe how such programs are used in a commercial product Diagnostica.