Concept
Dunning-Kruger Effect
overconfidence self-assessment cognitive-bias
People with low competence at a task tend to substantially overestimate their own ability at it, since the same lack of skill that produces poor performance also impairs the ability to accurately judge that performance. Eva van den Broek shares the origin story researchers David Dunning and Justin Kruger cite for what first drew their interest: a man who robbed banks in broad daylight rubbing lemon juice on his face, apparently believing — based on a misunderstanding of lemon juice's use as invisible ink — that it would make him invisible to security cameras. He was caught almost immediately and was reportedly genuinely baffled at being recognized.
A controlled demonstration that overconfidence intensifies specifically when it has social value. Economists Joel van der Weele and Peter Schwardmann gave participants an intelligence test and asked them to estimate their own score, both before and after receiving feedback. Most participants were overconfident by default, rating themselves above average. A subset of participants was told their self-estimate would be used by an "employer" figure to decide whether to select them, and predictably rated themselves as even more confident under that condition. Separately, some participants received false, artificially positive feedback regardless of their actual (sometimes poor) performance — those given inflated feedback became more overconfident and were subsequently more persuasive at convincing an evaluator of their intelligence, despite performing no better on the actual test. See Van der Weele Schwardmann Overconfidence-Signaling Study.
Practical implication: people who most need to guard against overconfidence — because they're weakest at a task, or because they stand to benefit from appearing more capable than they are — are systematically the worst-positioned to notice it in themselves, and inflated confidence can genuinely help convince others even when it has no basis in real skill.
A closely related, broader bias: illusory superiority (Episode 235). Louis Grenier cites surveys finding roughly 90% of people rate themselves smarter than average, and roughly 80% of surveyed UK prisoners rate themselves more caring and law-abiding than average — both statistically impossible at a population level. He connects this to why marketing "secrets" and hacks are so appealing: believing you've found a hidden trick lets you feel smarter than everyone who hasn't, and being told "there is no secret" is uncomfortable precisely because it removes that flattering belief (see Illusory Superiority Survey Stats).
The actual founding empirical study (Episode 278), not just its origin anecdote. David Dunning and Justin Kruger's 1999 Cornell paper (JPSP) tested participants on humor judgment, grammar, and logic, then asked them to rate their own performance. Participants in the bottom quartile were both bad at the task and badly overestimated their own ability — but participants in the top quartile underestimated their own relative standing too, apparently assuming others were roughly as skilled as they were. The bias, in other words, runs in both directions. Illustrated with the same lily-pad puzzle used elsewhere on this wiki for overconfidence (most people answer "day 24" for when a doubling-daily lake is half-covered; the correct answer is day 47). Chess grandmaster Magnus Carlsen's habit of harshly criticizing his own play even after winning is cited as a real-world instance of expert underestimation.
Discussed in
- Episode 227 — 227-tiny-nudges-that-can-drastically-improve-your-life
- Episode 235 — 235-louis-grenier-s-extremely-uncensored-take-on-marketing (illusory superiority survey stats)
- Episode 278 — 278-real-world-examples-of-cognitive-biases (the actual 1999 founding study; lily-pad puzzle; Magnus Carlsen example)