10 ms·

Hm, that relationship is also what you'd get if people were perfect at estimating their own skill while the proxy being used by the researchers was imperfect, i

• klodolph 8y ago
  Well, you would get broadly speaking a similar relationship. General error in the measurement would appear as a regression to the mean. However, Dunning and Kruger addressed this in their paper, directly, in section 4.1.3. The actual metric that the researchers measured was actual class rank and estimated class rank. Class rank is well-defined, and estimated class rank is also well-defined. That's not to say that class rank is a perfect choice of metric, but you are asking students to estimate something which can be exactly measured. The paper showed a difference in the size of the error between top and bottom quartile students that was larger than you would expect if it were just uniform error. When you think about it, this seems pretty damn obvious... if you are bad at something, you are also bad at knowing that you are bad at it. But if you are good at something, you have a more accurate understanding of your own skills. In short, if you're good at something, you probably know that you are good at it. This is what goes against the popular notion of the Dunning Kruger effect. Many people seem to use Dunning Kruger as some kind of twisted justification for impostor syndrome, where you are not permitted to think you are good at something because it would paradoxically imply that you are not good at it. I say this, just because I know too many people suffering from impostor syndrome, and this incorrect impression of Dunning Kruger seems to be a contributing factor.