Nick Barendregt, Department of Applied Mathematics, University of Colorado Boulder
Adaptive Decision Rules are Optimal in Simple Environments
Decision-making in uncertain environments often requires adaptive forms of evidence accumulation, but less is known about the decision rules needed to achieve optimal performance. While recent studies of decision models in stochastic and dynamic environments have resulted in several phenomenological models, such as the monotonically collapsing decision threshold of the “urgency-gating model” (UGM), we lack a general, normative description of decision rules and their relation to human decision-making. In this talk, we will study the prevalence of adaptive decision rules by developing a normative, Bayes-optimal framework for relatively simple two-alternative forced choice tasks. By allowing context variables, such as reward or difficulty of the decision, to vary in time, we find rich non-monotonic decision rules that vary throughout task parameter space. By comparing the performance of these strategies against simple heuristics, we show that these complex normative strategies significantly outperform alternative models such as the UGM. Finally, using subject data from the classic “tokens task”, we perform rigorous model fitting and comparison which suggests humans may use adaptive normative strategies in such tasks. These results provide testable hypotheses for experimentalists to validate in future psychophysics tasks and give insights into the complexities of human decision strategies.