A Parallel Self-Modifying Default Reasoning System

Engineering

Volume

Proceedings of the AAAI Conference on Artificial Intelligence, 5

Automated Reasoning

As a step in our efforts toward the study of real-time monitoring of the inferential process in reasoning systems, we have devised a method of representing knowledge for the purpose of default reasoning. A meta-level implementation that permits effective monitoring of the deductive process as it proceeds, providing information on the state of the answer procurement process, has been developed on the Parallel Inference System (PRISM) at the University of Maryland. Also described is an implementation in PROLOG (and to be incorporated in the above) of a learning feature used to calculate, for purposes of issuing default answers, the current depth of inference for a query from that obtained from similar queries posed earlier.