A CONCEPTUAL BASIS FOR EXPERT SYSTEMS METHODOLOGY

The application of knowledge engineering techniques to expert and consultative systems has tended to out-run the development of a conceptual basis for such applications. This paper presents some conceptual analysis, within the domain of knowledge engineering, of expertise, rule-following and, in particular, of competence, which is perhaps the crucial concept in the area. Examples are given from the computer fault-diagnostic system, CRIB, to support the argument. 1.0 Introduction 1.1 The 'conventional wisdom' of knowledge engineering [11 makes several assumptions concerning the application domain and the nature of expertise in the domain. Some are obvious, some not so. They include: 1. The best person to ask about doing a job is an expert. 2. An expert is good at his job, and better than other people. 3. The expert's domain knowledge can be extracted and represented on a computer. 4. The expert's heuristic skills can be represented by a set of rules. 5. Facts about the domain can be represented by a database of (usually) logical statements. 6. Different experts can come to agree through working with the computer system. 7. Experts may, over time, come to modify their methods of working to suit that of the computer. These seven points are representative of the beliefs and aspirations of workers in knowledge engineering, but I shall not claim that the list is exhaustive. Indeed, in a very young area of computer systems methodology, rapid change is inevitable, even at this fundamental level. 1.2 Point 1 is almost tautological; perhaps it is best considered as a definition of the word expert. But point 2 is more important when considering expertise in general. The acknowledged expert is so considered because he is good at his job. We cannot, of course, ignore the possibility of errors in judgement, but we can state with certainty that experts are normally good at their job-they are competent in their field. The notion of expertise would not get a hold at all unless this were so. 1.3 Hand-in-hand with the concept of expertise goes the concept of appropriate domain knowledge. Experts are expected and assumed to have more knowledge than other non-experts. This, I would argue, is because we need to find a basis for their displayed competence over and above the purely mechanical skills which experts also possess. For example, consider any lawyer, an acknowledged expert in all matters legal. (Note that expertise is also a relative …