META THEORY AS A TOOL FOR INTEGRATION AND CONTROL

Abstract

The use of meta knowledge and meta reasoning is well known and powerful approach in building of AI systems. Here we suggest to apply the meta approach for the integration of different resources into a computational system and for the control of inference in the integrated system. For this purpose, the system should have an explicit description of its parts, the ability to reason about these parts and to modify them. As an application of this approach, we have developed a Second Generation Expert System (SGES) architecture with introspective abilities. All modules and functions of the SGES are uniformly described as theories and tasks at a meta level. The proposed vehicle for implementation of such a system is REFLOG, a logic language with a framework for supporting reflective computations.