Reference: Musen, M. A. Languages for Knowledge Acquisition: Building and Extending Models. Stanford, 1989.
Abstract: Knowledge acquisition often is described as a process whereby expertise is "transferred" from the minds of application specialists to those of knowledge engineers, and thence to the knowledge bases of expert systems. This popular view, however, is misleading. Knowledge acquisition is a creative and inventive activity in which system builders generate new computational models of intelligent behavior (Winograd & Flores, 1986). In building an expert system, developers first construct a general model of the application task to be performed; that model then is validated and revised as necessary, as the experts and the knowledge engineers attempt to fit their evolving model to specific application problems. Knowledge acquisition thus can be viewed as comprising two phases: (1) building a general task model--that is, creating an INTENTION; followed by (2) establishing the content knowledge in the domain that corroborates the general model--that is, creating an EXTENSION (Addis, 1987). In this paper, I shall discuss the special nature of these two phases of knowledge acquisition, and shall argue for the use of knowledge-system development tools that are specialized for these two discrete aspects of the expert-system life-cycle.