Neural coding of computational factors affecting decision making

Abstract

We constantly need to make decisions that can result in rewards of different amounts with different probabilities and at different timing. To characterize the neural coding of such computational factors affecting value-based decision making, we have investigated how reward information processing is influenced by parameters such as reward magnitude, probability, delay, effort, and uncertainty using either fMRI in healthy humans or intracranial recordings in patients with epilepsy. We decomposed brain signals modulated by these computational factors, showing that prediction error (PE), salient PE, and uncertainty signals are computed in partially overlapping brain circuits and that both transient and sustained uncertainty signals coexist in the brain. When investigating the neural representation of primary and secondary rewards, we found both a common brain network, including the ventromedial prefrontal cortex and ventral striatum, and a functional organization of the orbitofrontal cortex according to reward type. Moreover, separate valuation systems were engaged for delay and effort costs when deciding between options. Finally, genetic variations in dopamine-related genes influenced the response of the reward system and may contribute to individual differences in reward-seeking behavior and in predisposition to neuropsychiatric disorders.