ENS, room Jaurès, 24 rue Lhomond, 75005 Paris
A crucial aspect of cognitive reasoning is the capacity to process information hierarchically. For example, imagine you want to prepare a dish you once enjoyed at a restaurant. You try an online recipe but the outcome falls short of expectations. You ask yourself “is it me or is this the wrong recipe?” Depending on your confidence in your cooking skills, you may try the recipe a few more times but if the results remain unsatisfactory, you may switch to another recipe. Behavioral studies have shown that humans reason about their failures by assessing their confidence after one or more attempts. However, the neural computations supporting this high-level reasoning strategy are not understood. We sought to characterize these computations in a non-human primate model of hierarchical decision-making.
We trained monkeys to perform a task in which they would face a negative outcome either because of misjudging the stimulus, or because of covert switches between two stimulus-response contingency rules. Monkeys, like humans, relied on their confidence to decide whether to attribute failures to self or to the occasional rule switches. They treated each failure as evidence for a covert switch, but did so rationally by accumulating evidence over multiple trials and by relying on the fact that it is less likely to misjudge an easy stimulus.
We then performed electrophysiology and perturbation experiments to investigate the neural computations and causal roots of hierarchical reasoning in the nervous system. We focused on the dorsal medial frontal cortex (DMFC) and anterior cingulate cortex (ACC) that were previously implicated in error monitoring and adaptive decision-making. Electrophysiological recordings indicated that these areas carry signals reflecting the confidence of animals in their decisions. Subsequent perturbation experiments revealed that ACC functions downstream of DMFC
and is directly and specifically involved in inferring covert rule switches. These results reveal the computational principles of hierarchical reasoning as implemented by cortical circuits. Our work extends previous foundational work on the neurobiology of decision-making in animal models to the general domain of hierarchical reasoning.