Probabilistic Similarity Networks

This work describes a new generation of expert systems—called normative expert systems. These systems have the potential to provide better decision support than do traditional expert systems in domains where the accurate management of uncertainty is important. This potential for improvement arises because people, including experts, make mistakes when they make decisions under uncertainty. That is, people often deviate from the rules of decision theory, which provides a set of rational principles or gold standards for how people should behave when reasoning or making decisions under uncertainty. Decision theory includes the rules of probability and the principle that a person should always choose the alternative that maximizes his expected utility.