Beyond the Steeper Curve: AI-Mediated Metacognitive Decoupling and the Limits of the Dunning-Kruger Metaphor
This addresses the issue of understanding AI's impact on human cognition and tool design, offering a more nuanced model than existing metaphors, though it is incremental in synthesizing evidence.
The paper tackles the problem that the Dunning-Kruger metaphor inadequately explains AI's effects on human performance, finding that LLM use improves task output but degrades metacognitive accuracy and flattens competence-confidence gradients.
The common claim that generative AI simply amplifies the Dunning-Kruger effect is too coarse to capture the available evidence. The clearest findings instead suggest that large language model (LLM) use can improve observable output and short-term task performance while degrading metacognitive accuracy and flattening the classic competence-confidence gradient across skill groups. This paper synthesizes evidence from human-AI interaction, learning research, and model evaluation, and proposes the working model of AI-mediated metacognitive decoupling: a widening gap among produced output, underlying understanding, calibration accuracy, and self-assessed ability. This four-variable account better explains overconfidence, over- and under-reliance, crutch effects, and weak transfer than the simpler metaphor of a uniformly steeper Dunning-Kruger curve. The paper concludes with implications for tool design, assessment, and knowledge work.