Reasoning with uncertainty in computer chess

摘要

This paper aims at an improvement of decision making under conditions of uncertainty. An overall analysis is given of how manifestations of uncertainty are dealt with in the field of computer chess. A new method of expressing uncertainty is presented which is done on the basis of a pair of point values associated with a weighting factor that indicates a preference between them. The reasoning process aiming at decisions among problem states associated with such a weighted pair is embedded in a traditional environment which requires point values. Essential components of this process are the overall (general) state of the critical position in terms of the degree of advantage and the competence of the system to judge the category of the domain-specific feature which causes the uncertainty. Finally, we present further improvements of the reasoning process which can be achieved when the requirement to back up point values is removed.