Abstract:

Substantial data support a temporal difference (TD) model of dopamine (DA) neuron activity in which the cells provide a global error signal for reinforcement learning. However, in certain circumstances, DA activity seems anomalous under the TD model, responding to non-rewarding stimuli. We address these anomalies by suggesting that DA cells multiplex information about reward bonuses, including Sutton's exploration bonuses and Ng et al's non-distorting shaping bonuses. We interpret this additional role for DA in terms of the unconditional attentional and psychomotor effects of dopamine, having the computational role of guiding exploration.

Publisher's Version