D&K ironically misengineered their tests and inadventently misconstrued their data due to floor and ceiling effects. If you ran the gamut of their tests against random noise you get similar results.

https://www.mcgill.ca/oss/article/critical-thinking/dunning-...

I posit that anyone who uses DK unironically is actually committing to the DK-paradox, something I'll leave you to define for yourself.