Taming Intuitive Predictions

A disciplined prediction must be pulled back toward the average in proportion to how weak the evidence actually is.

The mechanism is that System 1 substitutes evaluation for prediction. Shown that a child reads fluently far ahead of her age, we are impressed and predict an equally extreme future — a top grade-point average — as if reading precocity perfectly forecast college performance. But the correlation is modest, and a modest correlation cannot support an extreme prediction; the appropriate forecast is much closer to average. Intuition skips this correction entirely, producing predictions that are far too confident and far too extreme.

Kahneman lays out the corrective procedure explicitly. First, anchor on the baseline — the average outcome for the relevant class. Second, form the prediction your intuition suggests from the evidence. Third, estimate the correlation between the evidence and the outcome, and move from the baseline toward the intuitive prediction only by the fraction that correlation warrants. If the evidence is weakly diagnostic, the prediction should stay close to the average; only near-perfect evidence justifies an extreme forecast. This is, in effect, deliberately regressing your prediction toward the mean.

The consequence of skipping this discipline is a predictable bias toward overconfident extremes — overpredicting brilliant futures from precocious starts, catastrophic ones from early stumbles, and large effects from striking but unreliable signals. Kahneman notes that unbiased, regressive predictions will feel unsatisfying: they refuse to be impressed by a vivid case and will sometimes be embarrassingly moderate where a bold call would have looked prescient. But across many predictions, the disciplined, regressive approach is more accurate.

The applied takeaway is to temper bold forecasts drawn from striking evidence. When a candidate interviews brilliantly, a startup shows an explosive first month, or a student aces one test, resist predicting an equally extreme trajectory; ask how reliably the early signal actually predicts the long-run outcome, and pull your forecast toward the average accordingly. Strong impressions warrant strong predictions only when the evidence behind them is strongly diagnostic, which it rarely is.

Kahneman's deeper point is that there is a genuine trade-off between accuracy and the intuitive satisfaction of a confident, matching prediction. Regressive predictions are correct on average but will occasionally miss a true outlier, and they deny us the pleasure of the bold, vivid forecast. Choosing the disciplined method means accepting moderate, sometimes unexciting predictions as the price of being right more often — a deliberate victory of statistical reasoning over the intuitive pull to let the prediction be as extreme as the evidence that inspired it.