Logical Support for Design of Rule-Based Systems. Reliability and Quality Issues

In this paper selected logical foundations for design of reliable and efficient rule-based systems are investigated. The problems of logical analysis of theoretical properties of rule-based systems are considered. It is proposed to replace the verification of selected properties by assuring quality and reliability during the design stage of rule-based systems. Three basic properties, i.e. completeness, determinism and minimal number of rules are investigated and some ideas towards practical solutions are put forward. The formal base for this investigation is a logical analysis based on a single inference rule, i.e. backward dual resolution in generalized form. The design process itself is based on application of the presented reasoning rule, and a device for structuring the design process in the form of a special semantic tree ( -tree). The important characteristics, i.e. completeness, determinism and minimal number of rules of AI systems in context of reliability, safety and efficiency are discussed and are shown to cover several other issues. Theoretical results concerning the methodology of designing rule-based system assuring satisfaction of the above properties are given, and concluding remarks are enclosed. The proposed logical and methodological bases seem to be most suitable for rule-based control systems, i.e. one-step forward chaining ones.