“Reevaluating the Dunning-Kruger Effect: A Response to and Replication of Gignac & Zajenkowski2020”, Curtis S. Dunkel, Joseph Nedelec, Dimitri van der Linden2023 (IQ)⁠:

As applied to general intelligence, the Dunning-Kruger effect (DK) is the phenomenon in which individuals at the lower end of the intellectual ability distribution are more likely to overestimate their intelligence. In a recent article in Intelligence (Gignac & Zajenkowski2020) it was suggested that the DK is primarily a statistical artifact and, indeed, the application of more appropriate analyses led to a failure to replicate a statistically-significant effect.

When some of the limitations (namely sample representativeness) were addressed and the more appropriate statistical methods were used in the current study, our analyses:

illustrated a statistically-significant DK effect. However, the magnitude of the effect was minimal; bringing its meaningfulness into question…While the addition of the cubed term to the regression model also explained a statistically-significant amount of additional variance, the effect size was so small as to bring its meaningfulness into question. Gignac & Zajenkowski2020 recommended that changes in R² of 2% to 4% of additional variance explained is required to be indicative of substantial importance. In our analyses, the quadratic term increased the explained variance in self-assessed intelligence by 1.1% while the cubed term increased the model R² by <1% [ie. well into the ‘crud factor’ region]. Thus, while our analyses illustrated a statistically-significant DK effect, the magnitude of the effect appeared to be minimal.

In conclusion, it is recommended that the conditions that result in a statistically-significant DK be further explored.