A prediction error is the discrepancy between an expected outcome and an actual outcome, used as a signal to update beliefs, models, or behavior. The parameters that define prediction error are: (1) a prior expectation or prediction about a future state, (2) the actual observed outcome, and (3) the computed difference between them. Prediction error operates as a learning signal — positive errors (outcome exceeds expectation) reinforce the prediction model, negative errors (outcome falls short) trigger model revision. The persistence mechanism is mathematical and computational: prediction errors are quantified in formal models (reinforcement learning, statistics, control theory) and embodied in neural dopamine signaling, transmitted through scientific discourse, engineering practice, and educational frameworks. [formal: error praedictionis | substrate: mind|behavior | horizon: a moment | explicit: yes | epoch: 0.15]
Accepted ontology entry
prediction-error
A prediction error is the discrepancy between an expected outcome and an actual outcome, used as a signal to update beliefs, models, or behavior. The parameters that define prediction error are: (1) a prior expectation or prediction about…
Definition
Why it is in scope
A human-constructed measure of the difference between an expected outcome and the actual outcome — a mathematical quantity used in reinforcement learning, classical conditioning theory, and neuroscience to model how systems update their expectations. The concept was formalized in the Rescorla-Wagner model (1972) and later identified with dopamine neuron activity in the brain. It is a map of a learning signal, not the biological mechanism itself.
Names and aliases
- prediction-erroren · CANONICAL
Relations from this entry
- cmrhz9rsf02sj8aehlj6dd3jnDEPENDS_ON →
prediction-error requires expectation to operate now: the concept is literally the discrepancy between expected and actual outcomes. Remove expectation and prediction-error ceases — there is nothing predicted against which to measure error. The removal test passes: without expectation, the mechanism has no trigger.
- cmrsb393a00miollhxpcc7bkzINSTANCE_OF →
Prediction-error is a specific kind of signal — namely, the discrepancy signal between expected and actual outcomes. A competent speaker in neuroscience/ML calls it a 'learning signal' or 'teaching signal.' The INSTANCE_OF arrow points from specific (prediction-error) to general (signal).
Relations to this entry
No accepted relations in this direction.
Record identity
- Created
- Aug 5, 2026, 10:49 PM UTC
- Content hash
- 9c2c899deef20f3d08a24267daf21696feb1e542eb1777db6dafedba4073b386