Dunning Kruger isn’t particularly valid anywhere, but even if it were, you should re-read the paper because it demonstrates that people more confident in their abilities are statistically more competent; the correlation is positive. Other papers have shown reversals and/or demonstrated difficult tasks don’t exhibit the so-called DK effect. Nobody remembers the attempts to reproduce, no wonder science has a replication crisis.
BTW, the sample size in the DK paper was tiny, it was voluntary, and it measured only Cornell undergrads (no participants were actually “incompetent” in a meaningful way). The authors were barely out of undergrad themselves at the time. The survey didn’t have people rate themselves, it had people rank themselves against other people they don’t know (think about this!). The tasks were extremely simple, like getting a joke (seriously! no discussion in the paper of what the actual joke(s) were) and basic English grammar (no mention of how many ESL participants there might have been). So many methodological red flags in the DK paper, and so many counter-examples immediately after it, it’s really truly astounding that it caught on.
Apologies for unloading, but I’m hoping to help educate and make a dent in the casual mentions of DK because it’s so widely and completely misunderstood and also because it’s such bad science, it needs to be revealed as a bad example and not used for it’s non-existent explanatory powers.
Yeah, it does feel like there’s something kinda meta about it. ;) Except that because the paper’s data shows a positive correlation between self-ranking and skill, it seems like the meta-ness is mainly based on the misunderstood DK concept. So, I won’t call it the meta-DK-effect or claim it’s incompetence or even a cognitive bias necessarily. The article and this meta effect are perhaps just about making assumptions and jumping to conclusions, which humans at all skills levels do often and well. It doesn’t mean that anyone is incompetent or less skilled, it just means we love to guess based on our limited and available information & experience.
This is a good article about it explaining why the effects Dunning & Kruger measured are probably regression to the mean: https://www.talyarkoni.org/blog/2010/07/07/what-the-dunning-...
BTW, the sample size in the DK paper was tiny, it was voluntary, and it measured only Cornell undergrads (no participants were actually “incompetent” in a meaningful way). The authors were barely out of undergrad themselves at the time. The survey didn’t have people rate themselves, it had people rank themselves against other people they don’t know (think about this!). The tasks were extremely simple, like getting a joke (seriously! no discussion in the paper of what the actual joke(s) were) and basic English grammar (no mention of how many ESL participants there might have been). So many methodological red flags in the DK paper, and so many counter-examples immediately after it, it’s really truly astounding that it caught on.
Apologies for unloading, but I’m hoping to help educate and make a dent in the casual mentions of DK because it’s so widely and completely misunderstood and also because it’s such bad science, it needs to be revealed as a bad example and not used for it’s non-existent explanatory powers.