Rule-Based Expert Systems: The Mycin Experiments (1984) (www.shortliffe.net)

The 1984 book "Rule‑Based Expert Systems: The MYCIN Experiments" (eds. Buchanan & Shortliffe) is a comprehensive, retrospective collection of nearly a decade of research on MYCIN, Stanford’s seminal rule‑based medical expert system. The volume compiles detailed chapters on MYCIN’s rule form and inference engine, consultation and therapy algorithms, knowledge‑engineering practices, techniques for handling uncertainty (certainty factors, probabilistic models, and Dempster–Shafer theory), and tools for porting and building systems (EMYCIN). It also covers explanation generation, tutoring uses, hybrid representations (frames + rules), meta‑level knowledge, evaluation of system advice, and human‑engineering considerations—yielding a full experimental account rather than a theoretical treatise. For the AI/ML community the book is significant as a field‑defining empirical study of how to encode, maintain, explain, evaluate, and deploy expert knowledge. MYCIN’s experiments exposed strengths and limits of rule‑based architectures (e.g., modular rule bases, knowledge‑acquisition bottlenecks, and practical uncertainty handling), pioneered reusable inference engines, and foregrounded explainability and human factors—issues still central to modern ML. The collected lessons on knowledge representation, interpretability, evaluation methodology, and system usability remain valuable for current work in model transparency, hybrid symbolic–statistical systems, and trustworthy AI.