Generalization of the computational Dunning–Kruger account across tasks

Determine whether the computational conclusion that AI assistance removes the Dunning–Kruger effect generalizes across different reasoning task types.

Background

Prior work concluded that the Dunning–Kruger effect disappeared with AI assistance because a specified computational criterion was not met, although lower-performing AI users still overestimated more than higher-performing users. The present paper extends the investigation to four reasoning task types, but the quoted passage identifies the cross-task generalization of that computational conclusion as unresolved.