The Psychology of Being Confidently Wrong

In April 1995, a man named McArthur Wheeler walked into two banks in Pittsburgh in broad daylight and robbed them without a mask. His face, he had been persuaded, was protected by lemon juice, a liquid that works as invisible ink and would therefore make him invisible to security cameras. Police broadcast the footage that evening and arrested him within hours. He had reportedly tested the idea with a camera first and been pleased with the result. Wheeler is reported to have protested that he had worn the juice.

The story reached David Dunning, a psychologist at Cornell University, who suspected that Wheeler was not only wrong but unable to see how wrong he was. With his student Justin Kruger, Dunning tested the idea. In a paper published in 1999 under the title Unskilled and Unaware of It, the two asked Cornell undergraduates to take tests of humour, grammar and logical reasoning, and then to estimate how well they had done. Those who scored in the bottom quarter judged themselves to be far above where they stood, while the strongest performers slightly underrated themselves. The authors proposed a double burden. The skills needed to produce a right answer are the same skills needed to recognise one, so the person who lacks them is deprived of both the answer and the means of noticing its absence.

The finding travelled far beyond the journal. It became a standard explanation for reckless drivers, confident investors and loud experts on every subject, and self help books and management seminars borrowed the label freely. A curve showing confidence rising to a peak of ignorance, collapsing, and then slowly climbing again circulated online as proof, although it differs from the bar charts the authors actually published. The popularity of the idea is itself suspicious, since an explanation that flatters the reader by describing other people is rarely examined with care.

Then came the statisticians. Edward Nuhfer and colleagues, writing from 2016 onward, argued that much of the famous pattern could be produced by noise alone. When people estimate their own rank with error, and when test scores are themselves measured with error, the lowest scorers will appear to overestimate and the highest will appear to underestimate, because there is nowhere else for their mistakes to fall. Random numbers, they argued, can imitate the result.

The popularity of the idea is itself suspicious, since an explanation that flatters the reader by describing other people is rarely examined with care.

Defenders reply that noise cannot be the whole story. In the original paper, a short training session in logical reasoning made the weakest participants more accurate about their own standing, which fits a real link between skill and self knowledge. Yet this answers the statistical objection only in part. A theory about mechanism and a critique about measurement can both be true at once, and the data seem unable to separate them cleanly.

The dispute leaves an uncomfortable regress. Every judgement about competence, including the judgement that a study of competence is sound, depends on someone's sense of how well they can see. Wheeler tested his theory on himself and the world corrected him within an afternoon. Most people never receive so swift a correction, and those who do not cannot be sure whether their confidence is earned, inflated, or merely the shadow of a number drawn at random.