Abstract
Decision-making is ubiquitous in our everyday lives, from selecting between food options at a meal to the cognitively effortful choices of an air traffic controller. Intuitively, we prioritize sensory evidence linked to desirable and rewarding outcomes, but have diminished focus on sensory evidence when tasks are cognitively effortful and offer little reward. Prominent perceptual decision-making theories view decision deliberation through an information processing perspective, where sensory evidence is always acted upon in the same way during deliberation. Yet these models lack a unifying principle that governs how deliberation is modulated across different experimental conditions, requiring model refitting for each new context. Here we propose a novel theoretical framework where deliberation is a controllable process governed by desirability and cognitive effort. We used classic control theory, linear quadratic regulator, to model decision-making. From this theory, evidence accumulation (integration) and urgency act as complementary and interacting mechanisms to drive a decision. Several hallmark features of decision-making emerged from our model, including skewed response times, the speed-accuracy tradeoff, and Hick’s law. Further, low-dimensional neural dynamics from recent neural recordings in monkeys mapped onto the urgency and evidence accumulation components of the model. Taken together, our results provide compelling support for the notion that deliberation is a controllable process that is tuned by task demands and biologically plausible objectives of obtaining desirable outcomes with minimal cognitive effort.
