The Effects of Total Sleep Deprivation on Bayesian Updating

Recent evidence suggests that nearly 25% of U.S. adults (47 million) suffer from some level of sleep deprivation. The impact of this sleep deprivation on the U.S. economy includes direct medical expenses related to sleep deprivation and related disorders, the cost of accidents, and the cost of reduced worker productivity. Sleep research has examined the effects of sleep deprivation on a number of performance measures, but the effects of sleep deprivation on decision-making under uncertainty are largely unknown. In this article, subjects perform a decision task (Grether, 1980) in both a well-rested and experimentally sleep-deprived state. The experimental task allows us to explore the extent to which subjects weight prior odds versus new evidence (i.e., information) when forming subjective (posterior) beliefs of a particular event. Wellrested subjects display a tendency to overweight the evidence in forming subjective posterior probability estimates, which is inconsistent with Bayes rule but possibly consistent with use of a ‘representativeness’ heuristic. In his original Bayes rule experiment, Grether (1980) also found that typical student-subjects overweighted the evidence relative to the prior odds in making posterior assessments. Ironically, behavior following sleep-deprivation is more consistent with the use of Bayes rule, because this treatment significantly reduces the (over)weight that subjects place on the new evidence. Because choice accuracy is not significantly affected by sleep deprivation, the significant difference in estimated decision-model parameters may indicate that the brain compensates under adversity in certain risky choice decision environments.