Development of expert systems

In Chapter 43 the incorporation of expertise and experience in data analysis by means of expert systems is described. The knowledge acquisition bottleneck and the brittleness of domain expertise are, however, the major drawbacks in the development of expert systems. This has stimulated research on alternative techniques. Artificial neural networks (ANN) were first developed as a model of the human brain structure. The computerized version turned out to be suitable for performing tasks that are considered to be difficult to solve by classical techniques. 

Barratt, M.D., et al., Development of expert system rulebase for identifying contact allergens. Toxicology in vitro, 8, 1053, 1994. 

The primary differences, then, between development of expert systems and more traditional software engineering are found in steps one and two, above. First, the problems chosen will involve symbolic reasoning, and will require the transfer of expertise from experts to a knowledge base. Second, rapid prototyping, the "try it and see how it works, then fix it or throw it away" approach will play an important role in system development. 

The only phase of development of expert systems that I will say any more about is the first, and in many ways the most crucial, step for those who are contemplating building expert systems for the first time. How do you go about selecting an appropriate application Here are the basic criteria ... 

The development of expert systems need not be costly. There are several expert system shells commercially available, and an expert in artificial intelligence (AI) is no longer needed to program them. Simple IF-THEN rules can easily be programmed in more commonly used languages like FORTRAN and BASIC. In fact, the McDonnell Aircraft expert system referred to earilier was programmed in FORTRAN. Cost depends on the number of rules and... 

All of the above argues for development of expert system approaches for 3D molecular structure determination, and these are expected over 5-10 years, at least for the most studied systems proteins and nucleic acids. 

Expert systems use this approach for problem solving by separating technical aspects clearly from the domain aspects. Consequently, the domain experts are seldom development experts who create an expert system. However, developers of expert systems are able to create a reusable shell that allows domain experts to enter knowledge in an understandable fashion. 

Conventional algorithms, commercial software, or solution approaches are not available. Scientific problems often arise from new analytical results and insights. Many of the problems to be solved resulting from new findings are rarely covered by literature, not to mention software algorithms. Rule-based systems are one of the alternatives for an easier development of expert systems. 

In the following sections, we will examine some of the programming languages that are of importance for the development of expert systems. 

Issues for development of expert systems can be divided into eight general categories, which parallel the phases of life cycle management. 

A secondary objective of this project is to explore the development of expert systems in a mainframe platform in the EPA network. Intelligent system components prototyped in the microcomputer could be implemented on a mainframe computer. This is an alternative to the expense of mainframe development that has deterred the implementation of expert systems for this platform at EPA. The relatively high cost of expert system development tools for mainframe computers could be obviated by transferring expert system shells developed by other Government agencies to the EPA. Some of the choices are CLIPS, developed by the National Aeronautics and Space Administration, and LES (Lockheed Expert System), developed by Lockheed within Department of Defense projects. 

Developers of expert systems for the EPA must recognize important Agency considerations, such as the varying nature of the problems addressed, a staff with diverse experience, decentralization of tasks, computer resource constraints, geographical distance, and high personnel turnover. 

In spite of the many perceived benefits, the development of expert systems per se over recent years has been surprisingly slower than one would expect. One possible explanation for this is that when the systems were first introduced, their capabilities were overestimated and they were seen as the panacea to all formulation problems. This was obviously not the case, but as a result, the systems are viewed with some degree of scepticism. Several reviews on the issues and limitations with the development of an expert system have been published (Dewar 1989 Tinsley 1992 Rees 1996). 

The principle conclusion is that the development of "expert" systems and the necessary audit tools for verification and validation of those systems is still in infancy. There are a number of interesting concepts under development which bear promise but which will probably see their first application within critical space and defense applications. Tools useful to those of us interested in more generalized risk and reliability analyses are probably at least two to three years away. 

The development of expert systems is of interest, for example, for interpretation of spectra. There exist expert systems for... 

With the development of expert systems, it is possible to think of designing an "advisor" into computer-based instruction - an advisor which will monitor the learner s Activities, intervene with advice when appropriate, answer questions about instructional management, and serve other instructional management functions. But what are the rules which should govern such an advisor And what instructional management activities are best left to a human These are important areas for work in the next decade. 

Dr. Marlene Jones is a senior researcher in Artificial Intelligence in the Alberta Research Council s Department of Advanced Technologies and an adjunct professor in the Department of Computational Science at the University of Saskatchewan. Prior to joining the Alberta Research Council in January 1987, Dr. Jones was a tenured Associate Professor at the University of Waterloo. Her qualifications include a PhD (Computer Science) from the University of Toronto and an MEd (location of Exceptional Children) from the University of Saskatchewan. Her research experience expands mote than a decade, with the last 8 years focussed upon research in the area of AI and Education. Dr Jones recent research projects include the development of expert systems for educational diagnosis, expert environments for curriculum and course development, and user modelling. 

The 1987 NRC report Agenda for Advancing Electrochemical Corrosion Science and Technology focused on the development of expert systems to provide solutions to organizations with corrosion problems. The presumption was that improvements in computer technology would allow the capture of knowledge in a way that would make it easily accessible to users (National Academy Press,... 

Development of expert systems for the diagnosis and treatment of diseases. 

The application of risk and reliability techniques to acquire knowledge in the development of expert system for fault diagnosis... 

---