Hardly any psychological finding is quoted as readily as this one: those who know little overestimate themselves the most. The Dunning–Kruger effect has become a kind of universal explanation. You encounter it in comment sections, lectures, and discussions about politics and vaccinations—usually accompanied by a familiar curve showing a peak of overconfidence and a trough of disillusionment.
Incidentally, that curve does not come from the original study. And the effect itself is far less clear-cut than its popularity suggests. It is a good example of how a scientific finding changes once it leaves the academic literature. It becomes simpler, more definite, and above all more useful for arguments. But with every step, it loses something of what originally defined it.
What the Study Originally Showed
In 1999, Justin Kruger and David Dunning at Cornell University published a paper titled Unskilled and Unaware of It. Students completed tests on logic, grammar, and humour, then estimated how well they had performed compared with others.
The result was summarized in a chart. Participants were divided into four groups according to their actual performance. For each group, the chart showed how well they had actually done and how well they believed they had done. On average, the weakest group performed well below average but rated itself somewhat above average. The strongest group slightly underestimated itself.
The authors interpreted this psychologically. People who are not very capable in a particular area often also lack the ability to recognize their own mistakes. Incompetence therefore carries its own blindness with it. This is a plausible and compelling idea, and it made the paper one of the most cited works in social psychology.
The Statistical Problem
For several years now, however, there has been an important criticism—one aimed not at the idea itself, but at the calculation. The pattern in the famous chart arises to a large extent even when self-assessment and actual performance are only weakly related, without any psychological mechanism at all.
It is easy to think through. Suppose everyone rates themselves as roughly average, with a bit of random variation upward and downward. Then the self-assessment of the weakest performers will inevitably be higher than their actual performance, because they can hardly rate themselves as worse than the very bottom. And the self-assessment of the strongest performers will inevitably be lower. This produces exactly the familiar image, even though no one in the model is especially blind to their own weaknesses.
Several studies have demonstrated this using random data. If you simulate values containing no Dunning–Kruger mechanism whatsoever and analyse them in the same way, the typical pattern still appears. A statistical analysis in a psychology journal explicitly described the effect as an artefact that can be derived without a psychological explanation.
In 2020, Gilles Gignac and Marcin Zajenkowski went a step further and examined intelligence data using methods that avoid this problem. They found that people do, on average, overestimate themselves, but that the degree of misjudgment was similar across the entire range of performance. The weakest participants were therefore not substantially more blind than the others.
This criticism is not directed at Dunning and Kruger as researchers. The statistical pitfalls of such group comparisons received less attention in 1999 than they do today, and the methods have developed further. It is a normal process in science for a finding to be examined more closely later and qualified as a result. What is unusual is how little of this has reached public awareness.
What Remains
That does not mean the question has been settled. There are opposing views arguing that a smaller effect remains even after the statistics are corrected, especially in certain tasks and samples. The debate is open, and that is more honest than either extreme: the effect is neither proven beyond doubt nor completely disproven.
What can be said with some confidence is this: overall, people assess their own performance only moderately accurately, and many overestimate themselves somewhat. This applies to almost everyone, not just to the uninformed. The popular version—according to which the most ignorant people systematically believe themselves to be experts—is much less strongly supported by the data than the curve suggests.
It is interesting how the effect is used in everyday life. It almost always serves to describe other people. Those who invoke Dunning–Kruger rarely mean themselves. In this way, a finding about generally inaccurate self-assessment becomes a tool for labelling a group of people as particularly unreflective. That is a shift with little to do with the original research.
And there is an irony that is difficult to miss. An effect about the tendency to overestimate one’s own knowledge is often cited by people who have never read the study and do not know the statistical criticism. This is not a reproach, but an indication of how knowledge gets passed on: as a catchy formula, rather than as a finding with caveats.
In practical terms, this leads to a modest but useful attitude. Your own assessment is a hypothesis, not a measurement. Anyone who wants to know how good they are at something needs external feedback—ideally specific and repeated. This applies at the beginning of a learning journey just as much as after many years of experience. Experience does not protect us from misjudgment; it merely shifts the areas in which it occurs.
What remains is a thought for a quiet morning. Not every famous finding is wrong, but many are more uncertain than their fame suggests. It is worth pausing briefly when faced with an explanation that seems almost too compelling, and asking where it actually came from—and whether the curve you have in mind was ever measured that way.