FoMind
How long is the Yangtze River? Give a range you're 90% sure contains the true answer.
Many people write "5,000 to 6,000 km" and feel safe. The real answer is about 6,300 km, outside the range.
This isn't a one-off. Psychologists have asked large numbers of people questions like this: ranges people claimed to be 90% sure about often captured the true value only around half the time.
What overconfidence looks like
- Ranges too narrow: believing you know things more precisely than you do.
- Probabilities too extreme: saying 90% for things that happen only 70% of the time.
- Overrating yourself: believing you drive better or meet deadlines more reliably than most people.
On FoMind's stats page, overconfidence shows up as the high-probability end of your calibration curve sitting below the diagonal: for example, forecasts you gave 90% came true only 70% of the time.
This is where the interactive demo “90% confidence interval quiz” lives: in the FoMind app you can drag the controls and see results update live.
Test yourself below
Below are 10 questions. For each one, give a range you're 90% sure about. If you're well calibrated, you should capture about 9; if you only capture five or six, your ranges need to be wider.
How to guard against it when forecasting
- After writing a probability, ask: "In what situations would I be wrong?" Seriously list two or three, then see whether your number should move a little toward 50%.
- Be especially careful with anything above 90%: are you really willing to say "I'd be wrong only once in ten times"?
- Check your calibration curve regularly and correct yourself with real data, not feelings.
How to spot it in a review
If a forecast you gave a high probability didn't come true, look back: did you barely consider the other side? Did you turn "I think it's very likely" straight into 90%? If so, overconfidence may have played a part.
What feels like 90% is often closer to 70%: before giving a high probability, seriously think about how you could be wrong.