Neural circuit mechanisms of value-based decision-making and reinforcement learning

Despite groundbreaking progress, currently we still know preciously little about the biophysical and circuit mechanisms of valuation and reward-dependent plasticity underlying adaptive choice behavior. For instance, whereas phasic firing of dopamine neurons has long been ascribed to represent reward-prediction error (RPE), only recently has research begun to uncover the mechanism of how such a signal is computed at the circuit level. In this chapter, we will briefly review neuroscience experiments and mathematical models on reward-dependent adaptive choice behavior and then focus on a biologically plausible, reward-modulated Hebbian synaptic plasticity rule. We will show that a decision-making neural circuit endowed with this learning rule is capable of accounting for behavioral and neurophysiological observations in a variety of value-based decision-making tasks, including foraging, competitive games, and probabilistic inference. Looking forward, an outstanding challenge is to elucidate the distributed nature of reward-dependent processes across a large-scale brain system.