Dunning-Kruger Effect

Dunning-Kruger Effect describes a gap between perceived and actual competence that is largest at the low end of a skill: the same lack of expertise that produces bad answers also removes the ability to judge that they're bad. It isn't "stupid people are overconfident" — it's that competence and the meta-skill of evaluating competence draw on the same underlying knowledge, so when one is missing, so is the other.

The mechanism is easiest to see in a domain with clean feedback. A beginner chess player loses a piece to a fork and doesn't know it's a fork; a grandmaster loses a piece and immediately catalogs three ways it could have been avoided. The beginner's confidence is not dishonest — they genuinely cannot see the gap, because seeing the gap is itself the skill they lack. The famous U-shaped curve with a "valley of despair" is not in Kruger and Dunning's data, though — their measured self-assessments run roughly flat-to-rising across skill levels, and what shrinks with competence is the gap; the valley graph is an internet fabrication routinely misattributed to the paper.

The effect gets over-cited as a personal insult ("that's such Dunning-Kruger") when it's really a statement about calibration, not intelligence — the original study's biggest effect was on tasks like grammar and logic where feedback is rare, not a claim about IQ. It also has a well-known replication wrinkle: some of the statistical pattern is reproducible from pure regression to the mean plus noisy self-report, without needing a psychological mechanism at all — worth knowing before citing it as settled science.

For engineers, the practical corollary is to distrust your own confidence exactly where you have the least feedback loop — a codebase you just joined, a language you're new to — and to treat a newcomer's confident pronouncement as informative about their exposure, not their judgment.