When you say you're "90% sure" about something, are you right about 90% of the time?

Research consistently shows most people are NOT well-calibrated. When people say "100% sure," they're wrong about 20% of the time! When asked for ranges that "definitely" include the answer, people are wrong 40%+ of the time. The gap between confidence and accuracy is systematic.

We FEEL like our confidence matches reality because we remember confirmations and forget misses. When we're "90% sure" and right, we remember. When we're wrong, we find excuses. This selective memory makes us THINK we're calibrated when we're not.

Surprisingly, experts are often MORE overconfident than novices! More knowledge breeds more confidence—but not proportionally more accuracy. Doctors, lawyers, financial analysts—studies show they're all systematically overconfident in their predictions.

Try this: Write down 10 predictions with your confidence level. Check them later. Most people find their "90% sure" predictions are right only 70-80% of the time. Actually tracking your predictions is the only way to know if you're calibrated—and most people never do.

You're right that calibration varies by person! Some people ARE better calibrated than others. "Superforecasters" identified by Philip Tetlock achieve near-perfect calibration. But they're rare exceptions—and they worked hard to become calibrated. The DEFAULT human state is overconfidence.

Calibration also varies by domain. Weather forecasters are quite well-calibrated (they get immediate feedback). But political pundits, financial analysts, and medical diagnosticians show systematic overconfidence. Domains with clear, fast feedback tend to produce better calibration.

While variation exists, the AVERAGE person is significantly overconfident. Studies show "100% sure" means ~80% right. Without deliberate calibration training, almost everyone overestimates their accuracy. So "depends on the person" is true, but "most people are overconfident" is the better default assumption.

Well-calibrated people: (1) Track their predictions formally; (2) Seek feedback on outcomes; (3) Use specific probabilities, not vague words; (4) Consider reasons they might be wrong; (5) Update beliefs when evidence contradicts them. These habits are learnable—but most people don't practice them.

Correct! Research consistently shows systematic overconfidence. When people say "100% sure," they're wrong about 20% of the time. When asked for ranges that "definitely" contain the answer, they're wrong 40%+ of the time. The gap between stated confidence and actual accuracy is large and consistent.

Overconfidence may be evolutionarily adaptive—confident people are more persuasive, take more action, and project competence. But in modern decisions requiring accurate probability assessment, this inherited overconfidence becomes a systematic error. It's not stupidity—it's our default programming.

The good news: calibration is trainable! (1) Track predictions formally with specific probabilities; (2) Check outcomes and compare to your stated confidence; (3) Consider reasons you might be wrong before deciding; (4) Widen your confidence intervals; (5) Seek honest feedback. With practice, calibration improves dramatically.

Knowing you're likely overconfident changes decisions: Plan for more contingencies, widen time estimates, hold beliefs less tightly, be open to being wrong. Poor calibration causes planning failures, ignored warning signs, and overconfident bets. Good calibration = knowing what you don't know.