Bayesian versus heuristic-based choice under sleep restriction and suboptimal times of day

This paper examines how a commonly experienced cognitive state, sleepiness, impacts choice in a Bayesian task. We implement an ecologically valid sleep manipulation, and our main result is the increased use of a simple reinforcement-heuristic when sleepy. While use of this heuristic leads to errors in difficult decision environments where Bayes rule and reinforcement conflict, it does not harm choice accuracy in easier environments where Bayes rule and reinforcement coincide. Our results overall are consistent with the hypothesis that individuals make decisions in a relatively less deliberative manner under common adverse sleep states, which gives rise to a relative increase in automatic or heuristic-based decision making. While such heuristic-based decisions that are promoted by sleepiness need not be suboptimal in all contexts, our results show that the accuracy of a Bayesian assessment in more complex decision environments will likely suffer when sleepy.