Introduction: Knowledge acquisition as modeling

Nearly everyone in the field has heard the clichC that knowledge acquisition is the bottleneck in the development of knowledge-based systems. Developers frequently cite the so-called “knowledge-acquisition problem” as a reason to abandon knowledge-based systems in favor of other approaches, such as connectionism. This pessimistic view paints knowledge acquisition as the dark cloud overshadowing knowledge-based system development. From our perspective, knowledge acquisition is an opportunity rather than a problem. In fact, the most significant result of the widespread application of knowledge-based systems may be the attention it has focused on the knowledgeacquisition process. In our own work, we have sometimes found that the most important product of a specific knowledge-acquisition project is not the knowledge-based system, but rather the insight gained in the process of articulating, structuring, and critically evaluating a model of some domain. Likewise, we speculate that the fruits of knowledge-acquisition research efforts in the 1990s may turn out to be at least as significant as advances in implementation mechanisms were in the 1980s. Results of knowledge-acquisition research and practice have already been felt in areas as diverse as education, psychology, and engineering design. A recurring theme among the diverse collection of papers found in this issue is that knowledge acquisition is a modeling process, not merely an exercise in “expertise transfer” or “knowledge extraction. ” In particular, modeling is purposive, that is, to be involved in modeling is necessarily to be engaged in using the model (in some particular setting) for particular reasons that together determine what should be modeled, how to model it, and what can be ignored.’ Together, the criteria of purpose and cost-effectiveness determine how additional pragmatic issues should be resolved, such as who the users of the model