The first step in the development of an expert system is the extraction and characterization of the knowledge and skills of an expert. This step is widely regarded as the major bottleneck in the system development process. To assist knowledge engineers and others who might be interested in the development of an expert system, I offer (1) a working classification of methods for extracting an expert's knowledge, (2) some ideas about the types of data that the methods yield, and (3) a set of criteria by which the methods can be compared relative to the needs of the system developer. The discussion highlights certain issues, including the contrast between the empirical approach taken by experimental psychologists and the formalism-oriented approach that is generally taken by cognitive scientists.

Does the United States have a 51st state called Australia? A superficial look at the artificial intelligence (AI) research being done here could give that impression. A look beneath the surface, though, indicates some fundamental differences and reveals a dynamic and rapidly expanding AI community. General awareness of the Australian AI research community has been growing slowly for some time. AI was once considered a bit esoteric -- the domain of an almost lunatic fringe- but the large government -backed programs overseas, as well as an appreciation of the significance of AI products and potential impact on the community, have led to a reassessment of this image and to concerted attempt to discover how Australia is to contribute to the world AI research effort and hoe the country is to benefit from it. What we have seen as result is not an incremental creep of AI awareness in Australia but a quantum leap with significant industry and government support. The first systematic study of the Australian AI effort was undertaken by the Australian Department of Science (DOS) in 1986. The study took as its base the long-running research report Artificial Intelligence in Australia (AIIA), produced by John Debenham (1986). The picture that emerged is interesting. AI researchers are well qualified, undertaking research at the leading edge in their fields, and have significant potential to develop further. The results of this study were published by DOS in the Handbook of Research and Researchers in Artificial Intelligence in Australia (Department of Science1986). This article is based on key findings from the study and on additional information gained through meeting and talking with researchers and research groups.

This article discusses several effective techniques for expert system knowledge acquisition based on the techniques that were successfully used to develop the Central Office Maintenance Printout Analysis and Suggestion System (COMPASS). Knowledge acquisition is not a science, and expert system developers and experts must tailor their methodologies to fit their situation and the people involved. Developers of future expert systems should find a description of proven knowledge-acquisition techniques and an account of the experience of the COMPASS project in applying these techniques to be useful in developing their own knowledge-acquisition procedures.

Hand's book is well written and well researched. The author has taken great care in presenting previous work in detail and has quoted the erlier literature when applicable. Nevertheless, the book fails in two respects.

We describe a computer program that implements an algorithm to verify the consistency and completeness of knowledge bases built for the Lockheed expert system (LES) shell. The algorithms described here are not specific to this particular shell and can be applied to many rule-based systems. The computer program, which we call CHECK, combines logical principles as well as specific information about the knowledge representation formalism of LES. The program checks both goal-driven and data-driven rules. CHECK identifies inconsistencies in the knowledge base by looking for redundant rules, conflicting rules, subsumed rules, unnecessary IF conditions, and circular rule chains. Checking for completeness is done by looking for unreferenced attribute values, illegal attribute values, dead-end IF conditions, dead-end goals and unreachable conclusions. These conditions can be used to suggest missing rules and gaps in the knowledge base. The program also generates a chart that shows the dependencies among the rules. CHECK can help the knowledge engineer detect many programming errors even before the knowledge base testing phase. It also helps detect gaps in the knowledge base testing phase. It also helps detect gaps in the knowledge base that the knowledge engineer and the expert have overlooked. A wide variety of knowledge bases have been analyzed using CHECK.

Review of Expert Micros

In order for next-generation expert systems to demonstrate the performance, robustness, flexibility, and learning ability of human experts, they will have to be based on cognitive models of expert human reasoning and learning. We call such next-generation systems cognitive expert systems. Research at the Artificial Intelligence Laboratory at the University of Connecticut is directed toward understanding the principles underlying cognitive expert systems and developing computer programs embodying those principles. The Causal Model Acquisition System (CMACS) learns causal models of physical mechanisms by understanding real-world natural language explanations of those mechanisms.

The annual conference of the Association for the Advancement of Artificial Intelligence (AAAI) is the premier U.S. gathering for artificial intelligence (AI) theoreticians and practitioners. On the commercial side, AAAI is the only event with a comprehensive exhibition that includes most significant U.S. vendors of AI products and services. In 1986 some 5100 people attended AAAI- a very good showing considering that the 1987 International Joint Conference on Artificial Intelligence (IJCAI) drew about the same number of people even with its substantial international support. The commercial exhibits at AAAI-86 (110 exhibitors; 80,000 square feet) gave us opportunity to take a snapshot of an industry in transition.

An important issue in achieving acceptance of computer systems used by the nonprogramming community is the ability to communicate with these systems in natural language. Often, a great deal of time in the design of any such system is devoted to the natural language front end. An obvious way to simplify this task is to provide a portable natural language front-end tool or facility that is sophisticated enough to allow for a reasonable variety of input; allows modification; and, yet, is easy to use. It allows for user input to be in sentence or nonsentence form or both, provides a detailed parse tree that the user can access, and also provides the facility to generate responses and save information.

In 1940, a 20-year-old science fiction fan from Brooklyn found that he was growing tired of stories that endlessly repeated the myths of Frankenstein and Faust: Robots were created and destroyed their creator; robots were created and destroyed their creator-ad nauseum. So he began writing robot stories of his own. "[They were] robot stories of a new variety," he recalls. The young writer's name, of course, was Isaac Asimov (1964), and the robot stories he began writing that year have become classics of science fiction, the standards by which others are judged.