Effects of feedback on the weights and subjective values in an information integration model

Abstract

Judgment processes were studied in an information integration task requiring subjects to rate the quality of hypothetical job applicants. Three types of information were systematically varied and presented in all possible pairwise combinations such that only two types of information were available to the subject for any one applicant. Following each rating, outcome feedback was given by indicating the judged quality of the person after 6 months on the job. For one group, feedback was determined by only one of three types of information presented. For a second group, feedback was determined by a joint function of two types of information. In both cases the relevant information had a greater effect following feedback. While the effect of irrelevant information decreased in some cases, subjects continued to use this information despite feedback. The results were interpreted in terms of parameter changes in Anderson's information integration theory. A linear average model was supported that assigned a weight to each type of information. In some cases feedback resulted in an increased weighting of the relevant dimensions and in others the result was a change in the subjective value of information.