In the early 1970s, the young MD-PhD student Edward H. Shortliffe started work on his dissertation within Stanford University's Department of Computer Science. His PhD thesis, completed in 1975, involved the development of a computer program called MYCIN, an early expert system that aimed to furnish advice concerning the diagnosis and treatment of bacterial infections. MYCIN immediately attracted wide attention: it matured into a much larger research enterprise at Stanford and spawned larger conversations about whether and how to apply computers to the problem of medical diagnosis. This paper recounts this early effort to computerize medical diagnosis and decision making. It pays particularly close attention to the interrelations among computing, authority, and trust. How could a physician using the system know that its advice was accurate and trustworthy? What did the developers of MYCIN do to make the system's reasoning comprehensible to its human users? If the system's advice could not be understood, what perceived implications would that have for the physician's status as the ultimate decision maker? Such concerns about trust, authority, and comprehensibility animated both the development of and the responses to MYCIN. As artificial intelligence and machine learning are increasingly integrated into modern clinical care, these concerns endure today.
