An Introductory Course in Expert Systems

Terry Bahill
Systems and Industrial Engineering
University of Arizona
Tucson, AZ 85721-0020, USA
terry@sie.arizona.edu
© 1998-2004 Bahill

Artificial Intelligence (AI) is an old field that has recently delivered a prolific child: expert systems. The field of expert systems takes AI out of the realm of "pie in the sky" basic research and puts it squarely in development.

Expert systems are computer systems that emulate humans. Like the human, they can reason and make recommendations in the face of incorrect and erroneous data. However, expert systems are not going to replace humans, they are merely going to assist them: their purpose is to increase human productivity.

The granddaddy of expert systems is Mycin developed at Stanford in the 1970's to help a physician diagnose infectious blood diseases. All the knowledge put into this system existed in writing somewhere. So, instead of using the expert system the physician could have dug the information out textbooks, if they had good indices and tables of contents, and the user had enough time. The advantage of the expert system is that it leads the user through the knowledge base. Each question asked by the expert system is based on prior answers. Thus, the user does not answer all possible questions, only the relevant ones. In the metaphor of the interactive textbook, the user skips irrelevant chapters, but not solitary relevant details buried in seemingly irrelevant chapters. By using the expert system the user is able to decrease search time and reduce the number of errors.

This talk will describe an introductory undergraduate course on how to make expert systems using commercially available knowledge engineering tools (shells) that run on personal computers. The students built expert systems with various degrees of success. Lessons the instructors learned from that experience and the resulting changes in the course are described.

References [38 and 49]. This lecture is suitable for general scientists. It requires an overhead projector and an IBM compatible personal computer with a floppy disk drive and a projection device. This talk takes one hour.