Propagation of Individual Bias through Group Judgment: Error in the Treatment of Asymmetrically Informative Signals

TLDR: The evidence that individuals systematically violate Bayes' theorem under certain conditions is surveyed and two experiments designed to test whether individuals follow Bayesian reasoning and whether groups are able to overcome biased individual information processing are reported.

Abstract: Group decision making is commonly used in juries, businesses, and in politics to increase the informational basis for a decision and to improve decision accuracy. Recent work on generalizing Condercet's jury theorem provides a compelling justification for using groups in this manner. But these theories rely on a model of the individual as an optimal Bayesian decision maker. Do groups effectively aggregate information when the individuals are the flawed, non-Bayesian decision makers that actually populate acting groups? We first survey the evidence that individuals systematically violate Bayes' theorem under certain conditions. We then report two experiments designed to test whether individuals follow Bayesian reasoning and whether groups are able to overcome biased individual information processing. The experiments show that under certain conditions, with extreme probabilities and with signals that vary in diagnositicity, that individual accuracy actually deteriorates as information increases. For certain problems, majority rule effectively aggregates individual information. For the most difficult problems, majority rule fails to attenuate individual bias. The implications of these findings for research on individual and group judgment are discussed.