Prediction Error Signal
A neural signal marking the mismatch between what the brain expected and what actually happened, crucial for updating beliefs and learning in reinforcement.
A prediction error signal is a neural response representing the quantitative mismatch between an expected outcome and the actual outcome received. In reinforcement learning frameworks, this signal drives learning by updating internal models of the world, scaling the magnitude of the discrepancy to guide future behavior. It is a fundamental mechanism in brain computation, distinct from simple surprise or pain responses.
What this means in real life
When you expect a friend to text back within an hour but they don't reply for a day, your brain registers that mismatch—the prediction error—which makes you pay closer attention to their response when it finally arrives and may shift how you predict their future responsiveness.
What it isn’t
It is not simply noticing something unexpected. A prediction error signal is the specific neural mechanism that encodes the difference between expectation and reality, not just conscious surprise or awareness that something went wrong.
Commonly misused online
Often conflated with 'being wrong' or 'making a mistake.' Online, people use it loosely to mean 'I predicted badly,' but the term refers to a measurable neural signal that the brain uses to learn, not just the fact of inaccuracy.