MYCIN

MYCIN was an early expert system developed at Stanford University during the 1970s to assist clinicians in diagnosing bacterial infections and recommending antibiotic treatment. It became one of the best-known demonstrations of medical artificial intelligence and helped popularize rule-based reasoning and uncertainty management in computer systems.

A collaborative project led by Edward H. Shortliffe, MYCIN was designed to interact with users in a question-and-answer format, collecting patient data through dialog and applying a large knowledge base of if-then rules. The system contained around 450 rules that linked clinical findings to likely organisms and therapeutic options.

Reasoning in MYCIN relied on an inference engine that used backward chaining from potential diagnoses and certainty factors to represent confidence in propositions. The certainty factor calculus allowed the system to combine the strength of multiple evidential rules. An explanation facility was included to justify conclusions and prescriptions.

In addition to classifying organisms, MYCIN recommended antibiotic therapy and dosage ranges tailored to the patient when possible; it also provided rationale for treatment choices and highlighted uncertainties. It asked targeted questions to obtain missing data and avoided making recommendations without sufficient evidence.

Impact and legacy: Although not adopted in routine clinical care, MYCIN influenced the development of medical expert systems and informed the design of later inference engines and user interfaces. It underscored advantages and limitations of rule-based approaches, demonstrated the importance of uncertainty handling, and inspired generations of researchers to explore AI in medicine and the emergence of shells like EMYCIN.