Experiences with an expert system for gold exploration in Botswana

This article describes initial experiences with an expert system in the scope of an exploration project in Botswana. A simple and easy-to-use expert system shell for personal computers was applied to construct an inference network for epigenetic gold deposits. The main objective, however, was not the automation of human intelligence but the structural analysis of domain knowledge for a target assessment. Aspects of human reasoning are discussed and compared with machine reasoning. Some actual problems concerning the inversion of rules under uncertainty, contrary evidences, and prior probabilities are discussed. Technical aspects like costs and schedule are included in the exploration model. Three different approaches for constructing an inference network are described.