Reference: Gruber, T. The Use of Formally-Represented Engineering Knowledge to Support Human Communication and Memory. KSL, March, 1990.
Abstract: What can Artificial Intelligence do for engineering? The standard answer is given without hesitation: "automate it!" This ambition underlies the paradigm of automated reasoning: that what AI can do is automate processes currently performed by thinking human beings, such as design. I question the dominance of this view, and propose to emphasize another goal for AI. AI's unique contribution to science is knowledge representation: techniques for stating human knowledge in a formal, explicit, and operational manner. The contribution is not domain knowledge itself or alternate representations of physical mechanisms, but representations of knowledge that underlies reasoning. Engineers don't need AI people to come up with new ideas to do engineering, and they don't need new representations to automate well- understood engineering procedures. They already know about mechanics and thermodynamics and analysis of beams. Engineers can program, too.
Full paper available as hqx, ps.