Our confidence consistently runs ahead of our accuracy. When people say they're 99% sure, they're wrong far more than 1% of the time. We overrate what we know, underrate what could go wrong, and feel certainty the evidence doesn't earn.
The overconfidence effect: the confidence we feel in our judgments routinely exceeds their actual accuracy — we are surer than we are right, especially the less we know.
Ask a roomful of people for ranges they're '90% confident' contain the true answer — the population of a country, a historical date — and the true value falls outside their range far more than 10% of the time. Their confidence and their accuracy don't match. Kahneman shows this everywhere: experts predicting markets, doctors certain of diagnoses, founders sure of success, all systematically more confident than their track records justify. Worse, confidence feels like evidence, so the certain person is more persuasive than the accurate one.
Two forces drive it. We build a coherent story from the little we know and mistake its smoothness for truth — 'what you see is all there is,' so the unknowns simply don't register. And we're blind to our own blind spots (a cousin of Dunning-Kruger): the gaps in our knowledge are exactly the gaps we can't see. The result is planning that ignores what could go wrong, forecasts stated as near-certainties, and decisions made without margin for being wrong.
Calibration is the cure. Widen your ranges until they actually feel too wide — that's usually about right. Attach real probabilities to predictions and, over time, check whether your '80% sure' calls come true 80% of the time. Run a pre-mortem: assume you're wrong and ask why. And treat anyone's loud certainty — including your own — as the weak evidence it is. Strong conviction is not the same as being correct; the goal is confidence that matches reality, not confidence that feels good.
It's behind failed plans, blown forecasts, and bad bets — because feeling certain is mistaken for being right, we skip the margin of safety and the disconfirming check that being honestly uncertain would demand.
It's not 'never be confident, doubt everything.' Useful action needs conviction, and chronic under-confidence is its own failure. The point is CALIBRATION — confidence that matches your real accuracy. The fix isn't less certainty across the board; it's certainty that's earned, with margin kept for the cases where you're wrong.
Lock this idea into memory with a 5-minute active-recall session — the science of spaced repetition, no signup.
Try this idea free →