|
|
|
 |
|
EXTERNAL SEMINAR
Peter Bossaerts
Computation and Neural Systems Program, California Institute of Technology
Wednesday 27 September 2006, 16:00
Seminar Room B10 (Basement)
Alexandra House, 17 Queen Square, London, WC1N 3AR
Risk Prediction and Risk Prediction Errors in the Human Brain
All decision-makers confront many forms of uncertainty when making adaptive choices. One form occurs when stimulus-reward associations are probabilistic and changing. Understanding how organisms deal with this source of uncertainty has been advanced by a convergence between reinforcement learning models and primate physiology, which demonstrated that the brain encodes a reward prediction error signal. However, organisms must also track the level of risk associated with the reward predictions, monitor the errors in the risk prediction, and update risk prediction in light of the errors. To date, it is not known whether, or how, the brain accomplishes this. This led us to hypothesize that the brain encodes a risk prediction signal and a risk prediction error signal. The focus will be on insula, because of its demonstrated role in tracking uncertainty. In a simple gambling task, the insula BOLD signal correlates significantly and without delay with the risk prediction error, and, with some delay, with the predicted risk.