Theory 034/Mind & Behaviour /Behavioural neuroscience
Habits Need a Surprise
Habits are said to form with repetition. Repetition without surprise may leave the behaviour goal-directed forever.
Habit formation requires a threshold in reward prediction error rather than repetition alone, so a behaviour repeated without surprise will not become habitual regardless of how often it is performed.
01The evidence
Habit formation is variable across individuals and contexts even with matched repetition, and the variability tracks how surprising the outcomes were.
02The tension
Repetition-based accounts predict that practice makes habitual. That predicts uniformity, which is not observed.
03The mechanism
Model-free learning updates value only when outcomes differ from expectation. If outcomes are fully predicted, the learning signal is zero and the behaviour remains under goal-directed control.
That makes prediction error the engine, and predicts that varying the reward should accelerate habit formation while constant reward should delay it.
04What this is not
Repetition models count trials. This counts surprise, which predicts variability at matched practice.
05Predictions
- P1Habit strength should correlate with cumulative prediction error.
- P2Variable reward should accelerate habit.
- P3Constant reward should leave behaviour goal-directed.
06Refutation
Habit strength tracking repetition at controlled prediction error would restore the repetition account.
07What it changes
It would change habit-formation interventions.