Metacognitive Feedback Cut Answer Offloading to an LLM Assistant Roughly in Half (N=704)

arXiv 2609.20143·medium signalXBlueskyLinkedInCopy link

arXiv 2609.20143 (submitted 17 Sep 2026) tests two interventions against deskilling in a preregistered online experiment with 704 participants, a 2x2 design plus a no-AI control, where the task was practicing fraction arithmetic with an LLM assistant that gave solutions only on explicit request, followed by an unaided test. Metacognitive feedback that makes the implications of offloading explicit reduced answer offloading (odds ratio 0.47) and improved test performance (odds ratio 1.51). An effort-based reward incentivizing less extensive assistance showed no evidence of affecting either outcome, which is the more interesting negative result for anyone designing assistant UX.