Self & Social Bias · Quick Test
Dunning-Kruger Effect
Do people who know less think they know more? The story is more complicated.
The experiment
Answer the five reasoning questions. Then, without seeing the results, say how many you think you got right.
What happened?
You took a short test and, before seeing the answers, estimated your own score. That’s what Justin Kruger and David Dunning asked students to do in 1999. What they saw: the people who did worst greatly overestimated their performance, and those who did best underestimated it a little. A single estimate, like yours, doesn’t show that pattern; it only appears when you compare lots of people.
Why does your brain do this?
The original explanation is that, to know whether an answer is right, you need the same skill the question is testing. Someone who doesn’t know the logic rule doesn’t notice their own mistake, so they rate themselves higher than they did. But part of the pattern also comes from two simpler effects: almost everyone thinks they’re a bit above average, and noisy estimates pull the extremes toward the middle.
When does this shortcut work well?
Estimating how you did without checking everything saves time and gives you the confidence to try new things. For most of the day, a rough sense of “I think that went well” is enough, and people who know more about a subject tend to judge themselves more accurately in it.
Where does this show up in everyday life?
In the exam that felt easy and came back with a low grade, in the first month of a course when everything seems simple, and in the conversation where someone weighs in on something they don’t really know. The best antidote is concrete feedback: grade the test, ask someone who knows the subject, and compare your estimate with the actual result.
Nerd Mode
Open Nerd Mode
In Kruger and Dunning’s Study 2 (1999), 45 Cornell University students answered 20 logic questions taken from a prep guide for the LSAT, the US law school admission test. The bottom quarter scored, on average, at the 12th percentile, but thought they were at the 62nd and that they had gotten 14.2 questions right, versus 9.6 in fact. The top quarter scored at the 86th percentile and placed themselves at the 68th; they thought they had gotten 14.0 right, versus 16.9.
This item has contested evidence, and the dispute is statistical. The classic chart splits people into four groups by score and compares the score with the self-assessment. Edward Nuhfer and colleagues (2016) showed, using random numbers, that this kind of chart produces a similar shape even when nobody is really assessing themselves: measurement noise and regression to the mean are enough. Add the tendency to think you’re above average, and that accounts for much of the picture.
Gilles Gignac and Marcin Zajenkowski (2020) tested the hypothesis a different way, with 929 people who took an intelligence test (Raven’s Matrices) and estimated their own intelligence. The relationship between the two was essentially a straight line, and the estimates weren’t less accurate among the lowest scorers, contrary to what the hypothesis predicts. Their conclusion: the phenomenon may exist, but it’s much smaller than usually claimed.
Kruger and Dunning were already discussing regression to the mean in 1999 and tried to rule it out in Studies 3 and 4, and the debate goes on. One thing is certain: the famous “Mount Stupid” chart that circulates online isn’t in the paper. And the effect was never about “dumb” people: it’s about how anyone misjudges themselves in subjects they haven’t mastered yet.
Technical terms
- Self-assessment
- Overconfidence
Discovery
Described by Justin Kruger e David Dunning in 1999.
Level of evidence
Contested evidence
Major replications have failed or there is a strong alternative explanation.
Basis: Gignac, G. E.; Zajenkowski, M. (2020). The Dunning-Kruger effect is (mostly) a statistical artefact: Valid approaches to testing the hypothesis with individual differences data.
References
- Original studyKruger, J.; Dunning, D. (1999). Unskilled and unaware of it: How difficulties in recognizing one's own incompetence lead to inflated self-assessments. Journal of Personality and Social Psychology, 77(6), 1121–1134.
- Original studyNuhfer, E.; Cogan, C.; Fleischer, S.; Gaze, E.; Wirth, K. (2016). Random number simulations reveal how random noise affects the measurements and graphical portrayals of self-assessed competency. Numeracy, 9(1), 4.
- ReplicationGignac, G. E.; Zajenkowski, M. (2020). The Dunning-Kruger effect is (mostly) a statistical artefact: Valid approaches to testing the hypothesis with individual differences data. Intelligence, 80, 101449.
Review
This text has not been reviewed by a volunteer yet. Want to help review it?
Found a mistake? Write to [email protected]