Knowledge Elicitation: Principles, Techniques and Applications

This book sets out to present a coherent set of perspectives on knowledge elicitation, with the particular object of demonstrating the contribution which can be made by psychologists, linguists and sociologists. It arose from a meeting of the British Computer Society's Human-Computer Interaction Specialist Group in February 1988. The book concentrates on knowledge elicitation, i.e. the extraction and representation of expert knowledge, rather than the coding of that knowledge. It is intended for those who are ' . . . currently, or about to be, actually involved in knowledge elicitation and to provide a single reference for those involved in other aspects of building expert sys t ems . . . ' Its readership is therefore designed to be either academics or practitioners in industry who need to know about the complexities and realistic objectives of knowledge elicitation. It is not intended to be an undergraduate primer, but a suitable reference book for an 'advanced undergraduate' or postgraduate. It would be almost impossible to provide a 'comprehensive' overview of knowledge elicitation. The book comprises seven chapters by different authors, including one by the editor, who has also written short introductions to each of the other six chapters. This provides different, albeit mainly UK-oriented, perspectives, but the book has been edited so that chapters cross-refer to each other. There is also, in places, a healthy disagreement between authors. There is a strong view in the community, as voiced in this book, that knowledge elicitation can be viewed as almost distinct from other stages in expert system building. There is also, however, the view that knowledge elicitation is part of a modelling process in the life cycle of a software system, and that knowledge engineers can learn a lot from software engineering. This view is not strongly represented in the book, and an almost unfair criticism of prototyping is presented. Rapid prototyping can be used as a very effective means of knowledge elicitation. It is the control of the prototyping and system development which is crucial, which is why it may be unwise to concentrate on knowledge elicitation in its own right. Each chapter presents a different view, describing some mix of theory, methods and applications. The core of the book is one chapter which provides a fairly substantial and well referenced overview of different methods. The book also covers task analysis, mediating representations, a description of the knowledge engineer as a 'mediator', and some practical 'down to earth' advice about how to organize knowledge elicitation. In parts the book suffers because it fails to state explicitly which methods have been tried and tested, and on what type, size and complexity of systems. One suspects that some of the methods are proposed rather than the results of actual experience, whereas others have clearly been used in some domains. Given the aims of the text it would have been helpful had this been spelt out more clearly. The book could not meet its rather grandiose objectives, but it makes a useful contribution to the literature on the subject. It could not be, and would not claim to be, definitive, but it does present views of knowledge and of knowledge elicitation from a number of perspectives. It is worth reading or dipping into, but the reader must bear in mind that knowledge elicitation for expert systems is a part of the system development process, and that there are other perspectives which are not described in detail here. A multi-author book is bound to have different styles. In this case, the styles vary from the analytical to that almost of popular literature--I found the latter annoying, and it detracted from the content. The presentation is clear, and the bibliography and index are quite good. The book is hardbound, it is not underpriced, and it is a typical example of the Ellis Horwood series on expert systems.