The Effects of Artificial Intelligence Assistance on Meta-Cognition in Problem Solving

ABSTRACT In recent years, artificial intelligence (AI) has become part of ordinary work and study, promising better and faster outputs. It remains unclear how AI assistance affects the metacognitive processes that accompany problem solving. Specifically, we examined how people monitor their progress and when they decide to stop and provide an answer. We compared participants' performance on the Missing Tan Task (MTT; Ackerman, 2023), a multiple-choice problem-solving task based on geometric shapes, with and without AI assistance. On objective measures, AI assistance improved accuracy and speed. On metacognitive measures, the classic within-participant curvilinear association between confidence and time flattened with AI assistance relative to answering without AI. The results suggest that people set their stopping rules less strategically when working with AI. Future research is called for to examine whether these effects depend on how much of the work AI assistance leaves for people to complete.