Caroline Hayes, Michael Fu, and Dan Gaines
Knowledge acquisition is a large bottleneck in the construction of complex intelligent problem solvers. Unfortunately, the construction of effective KA tools to ease this process is also a bottleneck. In this paper we attempt to present some guidelines that can be used to determine whether one would do best to adopt a manual KA strategy, a computer-assisted KA strategy, or a completely automated KA strategy. Three knowledge properties that are relevant to determining the applicable KA strateg.y are the generality of the knowledge, the ease with which the expert can express this knowledge, and the degree of control that the expert wishes to exert over that part of the knowledge base. We will examine knowledge in an automated planner, p3, and a designer’s assistant, SEDAR. The KA strategies that are most likely to be the advantageous for aquiring the various types of knowledge will be identified.