Expert Systems--Lawyers Beware!
Expert systems as a genre have not escaped criticism. Their supposed virtues are loudly trumpeted by the well-funded1 community of "knowledge engineers" who build them. The weight of uncritical literature in favor of expert systems in almost every conceivable domain deserves a counterpoise, to which this chapter will contribute.
Among the most powerful critics of expert systems and its parent domain, artificial intelligence, are H. L. Dreyfus and S. E. Dreyfus ( 1986), J. Weizenbaum ( 1976), and T. Winograd and C. F. Flores ( 1986). The Dreyfus brothers demonstrated five levels of skill, ranging from the novice to the expert, of which two, they argue on psychological grounds, are beyond the range of the machine. Weizenbaum questions the epistemological assumptions of the computer-oriented thinkers, and Winograd and Flores point to the solution in terms of human knowledge and expertise being rooted in the involvement of agents in an infinitely complex and unanalyzable reality of physical and social action. Their arguments are commended to anyone needing to develop resistance to the current deluge of sales literature on expert systems.
Undoubtedly, expert systems have roles to play in business, especially where complex routine decisions have to be made at an operational level ( Lee, 1985). They provide a means of delegating such decisions in the form of precise rules to be followed in a bureaucratic manner by a technician or a clerk. They can be especially useful in circumstances where there is too rapid a turnover of staff to make it economic to train the technicians or clerks in the relevant skills. But____________________