How citing “Dunning-Kruger” shows that you are in fact oblivious about your own incompetence.

“Dunning-Kruger”, the supposed effect that incompetent people are oblivious of their incompetence and thus overconfident about their true skill level, has become an internet epithet much used to end discussion in social media. But few if any of the hurlers of this “argumentum ad dunning-krugerii” have actually read the paper.

Oliver Beige
3 min readMay 5, 2020

One piece of evidence is the prevalence of this “Mt. Stupid” chart in that forum of scientific accuracy known as “the Internet”. The chart, often also including an “Impostor Valley”, has nothing to do with Kruger-Dunning (1999), the paper that gave rise to the meme.

Profs. Kruger and Dunning conducted a series of experiments on undergraduates at American universities that supposedly demonstrated how incompetent students overestimated their skill level while competent students underestimated theirs.

“Dunning-Kruger” ≠ Kruger-Dunning (1999)

The first thing that stands out here is that there is no timeline and no opportunity for learning: Profs. Kruger and Dunning just ran a series of independent skill exercises. No learning means no Mt. Stupid (or Mt. Overconfidence) and no Impostor Valley.

The next thing that stands out is that Profs. Kruger and Dunning imposed a peer group whose skill level on particular tasks is entirely opaque to each subject in their experiment. Students might know their GPA in relation to their classmates, but it’s not clear how that translates to acuity at a basic task.

So in the absence of a relatable peer group in the experiment the students likely substitute their own peer groups — the fellow students they socially interact with — for the whole class. In the stratified environment of US postsecondary education this is likely a peer group of similar academic achievement: the keeners, the gofers, the slackers.

With these peer groups in mind, the typical self-assessment curve they derive from their experiments, generally upward-sloping at a lower angle than the “true” skill curve, becomes a sober and defensible self-assessment against each subjects peer group.

Is this shifting peer groups counterhypothesis more consistent with the evidence than Kruger’s and Dunning’s incompetence hypothesis? This is a question the authors, both academics after all, had to address before submitting their paper, and reviewers plus editors bring up before giving their ok to publication.

Just to add insult to injury, the experiments also include tests where the “skill” to be demonstrated was agreement with a (truncated) expert panel on such subjective things as sophisticated humor. The scientific term for such an experimental setup is “horseshit”.

In summary, the oblivious incompetents in this whole mess were Profs. Kruger and Dunning, and in turn everyone who brings up this pseudoscientific piece of truthiness as a showstopper in a social media debate.

The view from Imposter Valley. Photo by John Fowler on Unsplash

“Dunning-Kruger” = Rozenblit-Keil (2002)

The candidate paper for the Mt. Stupid meme is a paper by Profs. Rozenblit and Keil from 2002, which observes the peak and dip in self-assessment in a longitudinal panel study on skill learning. So if someone calls you a “Dunning-Kruger”, just call them a “Rozenblit-Keil” and send them to this article.



Oliver Beige

I write about how technology shapes the world we live in.