Research · October 1974
MYCIN
The system advised on antibiotics for blood infections, held a dialogue with the physician, explained every step and attached a measure of confidence to its conclusion.
Why it matters
An expert system first acquired the three things it cannot be used without: explanation, handling of uncertainty, and rules kept separate from the mechanism.
Around 600 if-then rules, each with a certainty factor. The physician could ask why a question was being asked and get the chain of reasoning back. The inference engine was separated from the rule base and turned into EMYCIN, an empty shell reusable in other domains. MYCIN never entered clinical practice, partly over questions of liability, partly because it needed a terminal and ten minutes of a doctor's time.
Event record
- Event date
- October 1974
- Timeline date
- Event date
- Verification
- Sources gathered automatically · September 17, 2026
- Lines
- ID
- evt-0126
The thesis was defended at Stanford in October 1974; the book form appeared in 1976 from American Elsevier. The record anchors on the work, not the book.
Records that link to this one
- Extends A blinded evaluation of MYCIN's advice
The evaluation is of the system described in the 1974 thesis.
Antimicrobial selection by a computer: a blinded evaluation by infectious diseases experts - Related R1 and XCON at DEC
The same rule architecture, carried from medicine to hardware configuration.
- Related Bayesian networks
MYCIN's certainty factors are replaced by a calculation with a probabilistic justification.
- Contrasts Probabilistic Reasoning in Intelligent Systems
It offers a justified replacement for the certainty factors of expert systems.
Probabilistic Reasoning in Intelligent Systems: Networks of Plausible Inference - Related Ledley and Lusted: logic and probability of diagnosis
Both records concern a machine advising a physician on diagnosis: in 1959 as a mathematical formulation with probability and value theory, in 1974 as a program with rules and certainty factors. The 1975 MYCIN paper read does not cite Ledley and Lusted.
- Contrasts CASNET: a glaucoma model before ophthalmologists
The CASNET authors themselves contrast their model, which reasons through a causal network of disease states, with 'MYCIN's purely normative or rule-based approach'.
A Model-Based Consultation System for the Long-Term Management of Glaucoma (S. M. Weiss, C. A. Kulikowski, A. Safir), IJCAI-77, vol. 2, 826-832 - Builds on PUFF: an expert system in a hospital's daily work
According to the 1982 report, PUFF was built with EMYCIN, a generalization of MYCIN that keeps its rule interpreter, explanation and knowledge acquisition modules.
PUFF: An Expert System for Interpretation of Pulmonary Function Data (Janice S. Aikins, John C. Kunz, Edward H. Shortliffe, Robert J. Fallat), Stanford report STAN-CS-82-931 / HPP-82-13