Cognitive Biases · CB-15
People's subjective confidence in their own judgments is, on average, reliably and measurably higher than the objective accuracy of those judgments.
A well-documented bias in which a person's confidence in their own answers, predictions, or abilities systematically exceeds the actual accuracy of those judgments — appearing across domains from trivia questions to financial forecasts to professional expertise, and considered one of the most pervasive and consistently replicated biases in judgment research.
Extensively studied by Sarah Lichtenstein, Baruch Fischhoff, and colleagues starting in the 1970s, using calibration experiments where people asked to state '90% confidence' intervals for factual answers were correct far less than 90% of the time.
The Mechanism
Stated confidence vs. actual accuracy — the calibration gap
The gap between the two lines is the overconfidence effect itself — across decades of calibration studies, people's '90% confident' answers have been correct closer to 70-80% of the time on average, a consistent, measurable overstatement of their own accuracy.
01 · IT'S STRONGEST FOR DIFFICULT QUESTIONS, WEAKER (SOMETIMES REVERSED) FOR EASY ONES
The 'hard-easy effect' is a well-documented nuance
Calibration research consistently finds overconfidence is most pronounced on genuinely difficult questions, while on very easy questions people are sometimes slightly underconfident — meaning the effect isn't uniform across all judgment types, but concentrates specifically where uncertainty is highest and feedback is scarcest.
02 · IT'S DISTINCT FROM DUNNING-KRUGER, THOUGH RELATED
Overconfidence appears even among genuine experts, not just the unskilled
Unlike Dunning-Kruger (which specifically concerns low performers overestimating their own relative standing), the general overconfidence effect has been documented even among genuine domain experts — professional forecasters, doctors, and financial analysts have all shown measurable overconfidence in calibration studies despite real expertise.
03 · IT HAS DIRECT, COSTLY CONSEQUENCES IN FINANCIAL AND MEDICAL DECISION-MAKING
This isn't a purely academic curiosity
Overconfident financial forecasts have been linked in multiple studies to excessive trading and worse investment returns; overconfident diagnostic judgments in medicine have been linked to diagnostic errors — the effect has measurable real-world costs well beyond calibration experiments in a lab.
Where It Fails / Inversion
Where it fails / inversion
Some individuals and some domains show genuine underconfidence rather than overconfidence — particularly on very easy tasks, or among people with clinical depression or certain anxiety profiles — and assuming everyone is always overconfident risks pushing an already appropriately or under-confident person toward false bravado rather than accurate calibration.
How To Use It
Worked example · calibration training for forecasters
Organizations that train forecasters (intelligence analysts, financial strategists) using explicit calibration exercises — repeatedly stating confidence intervals on factual questions and getting immediate feedback on accuracy — measurably improve calibration over time, a well-documented intervention distinct from simply telling people to 'be more humble.'
How to use it
Before acting on a confidently stated prediction or judgment — your own or someone else's — ask for their track record of calibration on similar past judgments, not just their stated confidence level. A consistent pattern of overconfidence in the past is a much better predictor of future miscalibration than how certain someone sounds right now.
See Also