EXPERTS SYSTEMS: PERILS AND PROMISE.

The article examines the knowledge-based expert systems that achieve expert level performance using explicit representations of knowledge. It tries to capture some of the relevant experience with systems that clearly fit the used model, rather than trying to define precisely what is or is not an expert system. The emphasis here has been on the explicit representation of knowledge in a system, rather than system performance. Therefore, no matter how clever a C-Compiler is, it will not be called an expert system if it were not built in this style. Expert systems need not be rule based or contain a theorem prover and can be built in any language, although it is easier where language and environment readily support the conversion of user's intent into programs. Since building an expert system is a large software project, standard software engineering analyses are useful in understanding the scope and nature of the job. One can build expert systems for appropriate problems provided experts are available who are articulate, patient and committed to a project for at least the initial phases.