Life / Neuroscience & Psychology

Dopamine and reward prediction

Schultz, Dayan and Montague (1997) showed that dopamine neurons signal errors in predicted rewards, like the error term of temporal-difference learning in AI.

Recordings from primate dopamine neurons showed activity that tracks mistakes in predicting rewards, which fits the error term used in reinforcement-learning algorithms. The link tied brain learning to machine learning and shaped later work on addiction and psychiatric illness.