ABSTRACT
Some years ago it was stated that knowledge acquisition is a major bottleneck in the development of expert systems (Feigenbaum & McCorduck, 1984), and this view is still propounded. Although it is not unusual for knowledge engineers to experience great difficulty in getting experts to verbalize and formalize their knowledge, there is another fundamental problem. Often the overall objective of knowledge elicitation is to construct a computer-based system. In such a case the usefulness of the system, and therefore of the knowledge modeled in it, is assessed at least in part by the system’s users. Knowledge engineering therefore involves constructing models of knowledge that can be validated by experts and that prove useful to the intended users. This means that the knowledge elicitation process should be viewed in the context of system design, and not as mining out an expert’s knowledge. As Kidd says (1987), the process should take into account the different classes of users who are likely to use the system, their requirements, and the types of knowledge they bring to the problem-solving process.
