Human-centered knowledge acquisition: a structural learning theory approach

Abstract

This paper develops the application of structural learning theory (SLT) to support the knowledge engineer (KE) in the knowledge acquisition process and the development of expert systems. The underlying research focuses on the knowledge to elicit from skilled domain problem solvers, and the structure(i.e. form and type) of this knowledge using SLT to guide elicitation and interpretation. SLT explicitly models both declarative and procedural knowledge, while presuming an innate backward-chaining mechanism. Guidelines based on SLT allow knowledge engineers to concentrate on the human-centered knowledge of domain specific problem solvers. In fact, the SLT model presumes that skilled problem solvers do not automatically divulge all rules. This human-centered, needs-based approach provides a point of departure from previous knowledge acquisition methods and serves as a distinguishing feature of this knowledge acquisition method. Specifically grounded in SLT, distinct rule types are developed to be extracted from skilled domain problem solvers. Based on these rule types, guidelines are developed to aid the KEs in the acquisition process.