9632448 BUDESCU Many decisions under uncertainty depend on expert judgments, because suitable statistical information is lacking. Examples include military and international policy contexts, business forecasting, medical diagnosis, and earthquake prediction. Experts do not always agree with each other, nor do they always express themselves in ways that best reflect their underlying opinions. This research will develop new and improved procedures for eliciting and combining their judgments to yield the most accurate and diagnostic conclusions. The project will investigate factors that affect human judgment under conditions of incomplete knowledge. The research also will develop new analytic procedures for such judgments, which take into account the sources and nature of the random error components that are normally present. One line of the project will ask why, if people are generally truly overconfident, they appear systematically underconfident given specific procedures and methods of analysis. A second line will explore why, at least for some judgment tasks, the way in which people think about uncertain information depends on the type of response they are required to give. A third line will investigate how people learn to make probability estimates and how feedback affects those estimates. The goal of all three lines of research is to develop a cognitive model of human information processing under uncertainty that applies to a broad range of situations, and can be expressed mathematically. One application of the mathematical model to be developed is to serve as a basis for combining multiple subjective estimates. A fourth line of work will pursue this goal. This approach is novel, as no other approach to the issue of combining judgments has depended on a model of the individual judge. ***
