Overconfidence Bias
The tendency for our subjective confidence in our own judgments to exceed their objective accuracy — particularly visible when we attach probability ranges to predictions.
Sarah Lichtenstein, Baruch Fischhoff, and Lawrence Phillips' chapter in the 1982 Judgment Under Uncertainty volume reviewed two decades of calibration studies and found a consistent pattern: when ordinary people set 98% confidence intervals around their estimates of facts they don't know, the true value falls outside that interval roughly 30-40% of the time, not the expected 2%. Our intuitive confidence is dramatically too narrow. Overconfidence has at least three forms — overestimation (thinking you are better than you are), overplacement (thinking you are better than others), and overprecision (thinking your beliefs are more accurate than they are). The third is the one calibration studies measure, and it is the most resistant to feedback. Domain expertise helps in stable, fast-feedback environments like weather forecasting, where forecasters become well-calibrated within a few seasons. It barely helps — and can hurt — in long-feedback domains like geopolitics, investing, or strategic planning, where the loop between prediction and outcome is too slow to retrain confidence directly.
Frequently Asked Questions
Why are weather forecasters well-calibrated when most experts aren't?
Because their feedback loop is short, frequent, and unambiguous — every forecast is graded within hours or days, and the score is a clean number. Domains with slow, noisy, or contested feedback never give predictors enough corrective signal to tighten their confidence intervals.
What is the simplest exercise to expose your own overconfidence?
Take ten 90% confidence intervals on facts you can later look up — population of a country, year a film was released, distance between two cities. By the math, nine should contain the true value. Most untrained people get five to seven. The gap between expected and actual is your personal overconfidence number, and it usually shrinks the second time around.
Are the three forms of overconfidence really separate?
Yes — Don Moore and Paul Healy's 2008 review in Psychological Review showed that overestimation, overplacement, and overprecision can move independently and even oppose each other. Someone can be overprecise about their predictions while underestimating themselves relative to peers, depending on task difficulty and feedback structure.