Room 3 · The Winters · 1972

MYCIN, the Expert System That Diagnosed

MYCIN, the Expert System That Diagnosed (Edward Shortliffe and team · Stanford, 1972): Rules written by experts, not learned.

Developed at Stanford from the early 1970s, MYCIN recommended antibiotics for blood infections. It ran on about six hundred “if… then…” rules written together with doctors, and could explain why it reached each conclusion.

MYCIN was developed at Stanford University in the early and mid 1970s as the doctoral work of Edward Shortliffe, with Bruce Buchanan, Stanley Cohen and a team of physicians and computer scientists. Its name comes from the suffix shared by many antibiotics.

The problem it tackled was serious and specific: helping doctors choose treatment for patients with severe bacterial infections, such as infections of the blood, often before laboratory results were complete. MYCIN asked the physician a series of questions about the patient and the tests, and then suggested the likely organisms and an antibiotic regimen, with doses.

Its knowledge was not learned from data. It was written by hand as several hundred rules of the form "if these conditions hold, then this conclusion is suggested, with this degree of confidence." To handle uncertainty, the team invented certainty factors, numbers that expressed how strongly each piece of evidence supported a conclusion. MYCIN could also explain itself: asked "why?", it showed the rule it was trying to apply; asked "how?", it traced the chain of reasoning behind a conclusion.

In a blinded evaluation published in 1979, outside experts rated MYCIN's treatment choices as acceptable about as often as those of Stanford infectious disease specialists. Even so, MYCIN never treated real patients. Entering data took time, computers were not part of clinical routine, and questions about responsibility for a machine's advice remained unsolved.

Its most lasting product was a method. Remove the medical rules and what is left, an inference engine and an explanation system, became EMYCIN, a shell for building expert systems in other fields. That idea led to the commercial expert systems of the 1980s.

Ask your AI: when it gives advice, can it explain why, the way MYCIN could? Ask it to show its reasoning for one answer.

Did you know? In evaluations its recommendations compared well with those of specialists, but it was never used on real patients.

Source: www.shortliffe.net/Buchanan-Shortliffe-1984/MYCIN%20Book.htm

See it in the museum →

The Lighthill Report →

All of Room 3: The Winters

Cite this: Nazareno, I. (2026). MYCIN, the Expert System That Diagnosed (1972). New Mind Museum. Retrieved November 17, 2026, from https://newmindmuseum.com/museum/mycin-the-expert-system-that-diagnosed-1972/