MYCIN was an early medical expert system developed at Stanford University in the first half of the 1970s, growing out of Edward Shortliffe’s doctoral work. The full account is his book “Computer-Based Medical Consultations: MYCIN,” published by Elsevier in 1976, the date used for this entry; the underlying research ran in the early 1970s. The original chapters are freely available on Shortliffe’s own website.
MYCIN was designed to help physicians diagnose and treat bacterial infections of the blood and to recommend antibiotics at appropriate doses. It held its medical knowledge as a set of a few hundred “if-then” rules supplied by human experts, and it asked the doctor questions, then chained the rules backward from possible diagnoses to reach a recommendation. Because medical evidence is rarely certain, MYCIN introduced “certainty factors,” numbers attached to rules and conclusions that let it reason with partial confidence rather than yes-or-no facts.
Studies suggested MYCIN’s recommendations were as good as those of some infectious-disease specialists, yet it was never used in routine clinical practice. The obstacles were practical and legal as much as technical: the effort to enter and maintain the knowledge, the slow computers of the day, and unresolved questions about responsibility when software advises on patient care.
Why business readers should care: MYCIN is the textbook example of an expert system, the dominant commercial form of AI in the 1970s and 1980s. It showed both the promise of capturing expert knowledge as rules and the recurring problem that such systems are expensive to build, hard to maintain, and brittle outside their narrow domain.