M. Mittenbühler: Optimal and efficient allocation of cognitive resources for decision making

Abstract
When choosing a course of action, humans must first decide how much to deliberate about their choice and what to deliberate on. This type of meta-control is commonly thought of as a cost-benefit analysis, weighing an increase in accuracy gained through deliberation against the increase in cognitive effort. However, the computational costs that performing this analysis itself incurs seem to defeat its own purpose. Here, I present how tools from variational inference can be leveraged to (1) define the optimal information-theoretic trade-off between accuracy and cognitive effort in decision making and (2) optimize this trade-off implicitly, i.e., without the need to estimate or explicitly represent any of the involved quantities. I will show how this optimization predicts key behavioral effects in conflict adaption and task switching paradigms, specifically list-wide and item-specific proportion congruency and switch effects.