Brian R. Gaines and Mildred L. G. Shaw
Knowledge
Science Institute
University of Calgary
Alberta, Canada T2N 1N4
Knowledge acquisition research supports the generation of knowledge-based systems through the development of principles, techniques, methodologies and tools. What differentiates knowledge-based system development from conventional system development is the emphasis on in-depth understanding and formalization of the relations between the conceptual structures underlying expert performance and the computational structures capable of emulating that performance. Personal construct psychology is a theory of individual and group psychological and social processes that has been used extensively in knowledge acquisition research to model the cognitive processes of human experts. The psychology takes a constructivist position appropriate to the modeling of human knowledge processes but develops this through the characterization of human conceptual structures in axiomatic terms that translate directly to computational form. In particular, there is a close correspondence between the intensional logics of knowledge, belief and action developed in personal construct psychology, and the intensional logics for formal knowledge representation developed in artificial intelligence research as term subsumption, or KL-ONE-like, systems. This paper gives an overview of personal construct psychology and its expression as an intensional logic describing the cognitive processes of anticipatory agents, and uses this to survey knowledge acquisition tools deriving from personal construct psychology.
Knowledge-based systems development is targeted on the emulation of human high-level skilled performance in a program running on a digital computer. Knowledge acquisition research supports the generation of such programs through the development of principles, techniques, methodologies and tools. Knowledge acquisition is not a monolithic process but, like all software engineering, draws on many sources of information in diverse forms, such as specifications, experience, principles, laws, observation, and so on, recorded in a variety of media. A major part of the task of the knowledge-based system designer is to collect this diverse material, organize it effectively, and develop from it structures suitable for computer implementation.
What distinguishes software engineering for knowledge-based systems from that for conventional systems is the relative emphasis on two aspects of the development process. First, the knowledge-based system approach emphasizes the use of human experts as the primary sources of information. Second, it emphasizes the use of knowledge representation schema as the primary basis of implementation. From an integrative viewpoint, one can see these as particular foci of attention, rather than radical changes in approach: human experts have always been an important source of information in system development; and knowledge representation schema may be regarded as the culmination of a trend towards the use of very high level, declarative specification languages. What has differentiated knowledge-based system development from conventional system development is the emphasis on in-depth understanding and formalization of the relations between the conceptual structures underlying expert performance and the computational structures capable of emulating that performance. That is the relationship between the psychology of the expert performer and the ontology of the computational emulator has been viewed as a research challenge on the one hand, and as the basis of supporting practical system development on the other.
We have been very careful in the wording of the preceding paragraphs to avoid presuppositions about the nature and status of "conceptual structures underlying expert performance" and about the intrinsic possibility and scope of "emulation of human high-level skilled performance." In particular, nothing is presupposed about the reality and location of conceptual structures, and neither is anything presupposed about the suitability of such structures for modeling or emulating any or all aspects of human expertise. These are deep issues important to future directions in research and practice relating to knowledge-based systems, and we and others have addressed them in depth elsewhere (Clancey, 1991; Clancey and Roschelle, 1991; Gaines, 1993; Gaines, Shaw and Woodward, 1993). This paper adopts a pragmatic and retrospective stance, reporting on what has been achieved in the past decade in knowledge acquisition research based on personal construct psychology.
Figure 1 summarizes the issues addressed above and gives an overview of the current state of the art described in this paper. At the top left is the person regarded as capable of expert performance of some task in some domain. At the top right is the computer regarded as capable of emulating the performance of that task in that domain. The arrow from person to computer indicates that an objective of knowledge acquisition research is to support the required expertise transfer from the person to the computer.
Figure 1 Psychological and computational foundations for theories, methodologies and tools supporting expertise transfer
Research on knowledge-based systems has emphasized the importance of understanding the conceptual structures underlying expertise, and of having knowledge representation schema that represent them as directly as possible. On the far left of Figure 1 is shown the psychological model of the person that has been the basis of the work surveyed in this paper, Kelly's (1955) personal construct psychology. This psychological model focuses on people as anticipatory systems, themselves developing conceptual models so as to better understand, predict and control a world. On the far right of Figure 1 is shown the ontological model of the computational knowledge representation that has been the basis of the work surveyed in this paper, Brachman's term subsumption model for KL-ONE-like schema (Brachman and Schmolze, 1985). This ontological model focuses on computers as anticipatory systems, using conceptual models to understand, predict and control a world. The arrow at the bottom of Figure 1 linking the psychological and computational models indicates that one significant aspect of the work reported in this paper is the unification of these two frameworks. There is a very direct relationship between the psychological conceptual structures presupposed in personal construct psychology and those implemented in term subsumption logics, and this is highly significant in supporting expertise transfer as a process of modeling the basis of human skilled performance in operational terms.
The central section of Figure 1 gives an overview of the theories, methodologies and tools presented in this paper. The central loop from person to computer shows: knowledge elicitation tools used to develop a psychological model of skilled performance through representations of the skills in terms of conceptual structures; knowledge modeling tools used to develop this to be a computational model of the skill in terms of logical structures; and knowledge structuring tools used to further develop this to be an operational model that will run within a particular computer environment.
At the very center of Figure 1 is the most important aspect of supporting expertise transfer effectively, that of providing feedback to the human expert for purposes of validating that the knowledge structures at each stage of the transfer properly represent the basis of the skilled performance, and for verifying them operationally as far as possible at every stage of development. It is an important side-effect of the emphasis on the human expert in knowledge-based system development that much of the functionality of the tools involved is targeted on supporting human understanding of the development process. This is again consistent with trends in the support of conventional software engineering such as tools for "requirements tracking," and knowledge-based system development can be seen within this framework as focusing on the way in which the requirements themselves develop and are refined through the very process of system development. That is, in the terms of the discussion above, the tools supporting "expertise transfer" are developing and refining a model of the basis of the expertise as a specification for a system to emulate it. Expertise is being transferred through the development of a model in terms of knowledge that provides a computational basis for emulating that expertise.
Abstract, 1 Introduction, 2 Personal Construct Psychology and its Applications, 3 Knowledge Representation in Personal Construct Psychology, 4 The Intensional Logic of Personal Construct Psychology, 5 The Repertory Grid, 6 Distance Measures, Conceptual Clustering and Induction, 7 Eliciting Conceptual Structures from Groups, 8 Tools Based on Personal Construct Psychology--KSS0, 9 Tools Based on Personal Construct Psychology--Aquinas 10 Tools Based on Personal Construct Psychology--KSSn, 11 Integration with Related Systems, 12 Integration of Knowledge Acquisition Methodologies, 13 Conclusions, References,