JOIJRNAL OF MATHEMATICAL PSYCHOLOGY: 2. 180-189 (1965) Decisions Based on Conflicting and Inaccurate Observations’

Subjects learned the accuracies of 8 cues in a series of 50 learning trials and then used pairs of these cues to predict which of two equally likely symbols occurred in each of 100 test trials. It was concluded that: (a) estimates of cues with very high or very low accuracies are better than estimates of cues with intermediate accuracies; (b) accu-rate cues are perceived more realistically than are inaccurate cues; (c) people tend to maximize expected payoff when faced with conflicting information in binary choice problems; (d) conformity pressures, i.e., the desire to agree with two reports, strongly interfere with maximization if there is uncertainty about the maximizing response, but conformity pressures exert little influence when there is no uncertainty about the maximizing response. Real life problems often require decisions before all the relevant information is gathered or reviewed. The information that is obtained often comes from a variety of sources, each of which provides inaccurate and incomplete cues. The experiment reported in this paper is an investigation of the way people use combinations of inaccu-rate sources in a simple choice situation. Information from a highly untrustworthy source is ordinarily nearly useless. But if the source is untrustworthy enough and the number of possibilities small enough, such information may be useful exactly because of its untrustworthiness. The largest effect of this kind arises when the world must be in one or the other of two states