Dopamine reward signal
At choice indifference between a safe reward and a gamble, the subjective reward value is equal for the two options, even when the mean physical amounts differ (left). The dopamine response is also equal, even when the amounts differ (right: dotted vs. solid lines). Despite their same responses, the neurons are sensitive to reward value, as their response increases with increasing value (right: different colours).
At choice indifference between a safe reward and a gamble, the subjective reward value is equal for the two options, even when the mean physical amounts differ (left). The dopamine response is also equal, even when the amounts differ (right: dotted vs. solid lines). Despite their same responses, the neurons are sensitive to reward value, as their response increases with increasing value (right: different colours).
Economic utility formalises the subjective nature of reward value mathematically. Meaningful reward neurons should code utility rather than physical amount. The dopamine reward prediction error (RPE) signal follows nonlinear utility with unpredicted reward (left) and gambles (right: stronger dopamine UPE response at steeper utility slope despite same physical RPE).
Economic utility formalises the subjective nature of reward value mathematically. Meaningful reward neurons should code utility rather than physical amount. The dopamine reward prediction error (RPE) signal follows nonlinear utility with unpredicted reward (left) and gambles (right: stronger dopamine UPE response at steeper utility slope despite same physical RPE).
According to the frequentist approach, probability reflects the frequency of past events (top left). Reinforcement learning can implement that approach (top right). Monkeys learn to estimate probabilities in choice between a fixed known reward probability and different probabilities indicated by new stimuli (bottom left). Dopamine responses to the new stimuli show an indiscriminate novelty response that declines with stimulus experience and is followed shortly later by a differential value response that develops with experience (bottom right).
According to the frequentist approach, probability reflects the frequency of past events (top left). Reinforcement learning can implement that approach (top right). Monkeys learn to estimate probabilities in choice between a fixed known reward probability and different probabilities indicated by new stimuli (bottom left). Dopamine responses to the new stimuli show an indiscriminate novelty response that declines with stimulus experience and is followed shortly later by a differential value response that develops with experience (bottom right).
Medium efficacy but high specificity (left). Optogenetic stimulation adds to the dopamine response to juice reward (top center), which enhances the dopamine response to a reward predicting stimulus, compared to juice alone (centre bottom) and enhances the choice probability for that stimulus (right).
Medium efficacy but high specificity (left). Optogenetic stimulation adds to the dopamine response to juice reward (top center), which enhances the dopamine response to a reward predicting stimulus, compared to juice alone (centre bottom) and enhances the choice probability for that stimulus (right).