The use of nonlinear, noncompensatory models in decision making.

Abstract

An important problem in decision making concerns finding the utility of a multidimensional stimulus. This has traditionally been done by assuming that total utility is a linear function of the attributes of the stimulus. In clinical decision making, the linear regression model has been used to predict and diagnose on the basis of multidimensio nal information as well as to approximate the clinician's own judgment. Other nonlinear, noncompensatory models are available for combining information. These models, called conjunctive and disjunctive, are approximated here by suitable nonlinear functions of utility. They are then shown to give a better fit to certain decision data than the linear model. The factors affecting the use of these models and their implications are discussed.