From Dart to Designworld: a chronicle of research on automated engineering in the Stanford Logic Group
Michael Genesereth · Artificial Intelligence · 1993
For those of us in the Stanford Logic Group, the Dart project marks the beginning of a long-term commitment to research on the use of computers in the service of engineering. Oddly enough, the project began not in engineering, but in medicine; and it was originally concerned with explanation, not diagnosis. When I arrived at Stanford in 1979, I assembled a small research group with financial support from a research contract belonging to Bruce Buchanan and Ed Feigenbaum. I asked Bruce how we might earn our keep, and he suggested that we look into ways of improving Mycin's explanation capability. Mycin was already world-renowned for its ability to explain its conclusions by citing the data and rules used to make those conclusions. The problem was that it was unable to explain why its rules were correct. We knew that physicians could supply rationale for many of Mycin's rules by reference to underlying physiological principles. So, our idea was to capture this physiological knowledge and produce a program that could use it to verify Mycin's rules. A trace of the verification could then be used to explain the rules.