Dunning–Kruger Effect
A pattern in which people with the least skill in a domain tend to overestimate their ability, while highly skilled performers are more likely to underestimate themselves relative to peers.
Justin Kruger and David Dunning's 1999 Journal of Personality and Social Psychology paper showed that the bottom quartile of performers on tests of grammar, logic, and humor rated themselves near the 60th percentile, while the top quartile estimated themselves slightly below where they actually scored. The interpretation Kruger and Dunning offered: the same skills required to perform a task are often the skills required to recognise good performance, so the unskilled lack the metacognitive tools to see how unskilled they are. The effect is real, but widely misreported — it is not a clean U-shape, and it does not say that beginners think they are experts. It is a calibration story about systematically compressed self-estimates at both ends. Subsequent work by Burson, Larrick, and Klayman showed the pattern partly reflects regression toward the mean and noisy self-assessment. The practical takeaway holds either way: external feedback and direct comparison with measured performance correct calibration faster than introspection does.
Frequently Asked Questions
Does Dunning–Kruger mean novices think they are experts?
Not in the original data. Bottom-quartile performers in Kruger and Dunning's studies rated themselves around the 60th percentile — overestimated, but still average, not expert. The popular cartoon of a confidence "Mount Stupid" exaggerates what the research actually showed.
How can someone tell if they are in the over-confident zone?
Calibration tests work better than introspection. Pick a domain, make specific predictions with confidence levels, score them against reality over time, and look at the gap. Anyone whose 80%-confident calls are right 50% of the time is mis-calibrated, regardless of how skilled they feel.
Is the effect just statistical regression to the mean?
Partly. Burson, Larrick, and Klayman's 2006 work showed regression and measurement noise account for some of the curve. But the core finding — that low performers lack metacognitive insight into their own performance — has held up across many replications, with the regression critique refining rather than overturning it.