Dunning-Kruger should go the way of the Stanford Prison Experiment. For those who may not know about the latter, it was also largely faked. Guards were coached

Dunning-Kruger should go the way of the Stanford Prison Experiment. For those who may not know about the latter, it was also largely faked. Guards were coached in how to act, the famous 'Jesus Christ, I'm burning up inside. I can't stand another night!' prisoner was a student who wanted to go study (which he was told he was allowed to, but wasn't) and so he faked a mental breakdown, and so on. Back to Dunning Kruger, it's largely a statistical trick. Wiki actually has a pretty decent section on this [1], alongside some nice examples. The gist of the trick is that the more random a sample is (in other words the more people are unable to accurately self evaluate) the stronger a Dunning-Kruger effect you'd show. The way Dunning-Kruger achieved the really beautiful perfectly linear graph people always reference is by having people self evaluate their ability to... determine a funny joke. And so broken was their test that they had to eliminate one of their 8 professional comedian judges, because he couldn't determine a funny joke, by the standards of their test. And the remaining 7's scores were judged, by the paper itself, as "moderately reliable." And I suspect this was intentional. The first thing they would have done is to plot a scatter graph of their results. And they would certainly have quickly realized the transform (by bucketing) and the scatter plot tell radically different stories (again the illustrations on Wiki make this more than clear), but only one of them is true! [1] - https://en.wikipedia.org/wiki/Dunning%E2%80%93Kruger_effect#Statistical https://en.wikipedia.org/wiki/Dunning%E2%80%93Kruger_effect#...