Students often don't notice what they lose by letting AI answer for them — would pointing that out change their habits?

This explores whether prompting students to notice their own thinking can get them to stop asking AI for answers, and whether they learn more as a result. The corpus has a direct answer, and it is yes. In a preregistered experiment with 704 people, feedback that pointed out the costs of offloading cut the number of times students asked an LLM for answers by half. It also raised their scores on a later test taken without the AI by 51% Can metacognitive feedback stop students from offloading to AI?. The surprise is in the comparison condition. Rewarding students for effort had no measurable effect on either outcome. Paying people to try harder didn't help. Showing them what their habit was costing them did.

Why would awareness work when incentives don't? Part of the answer is that students often can't tell they're losing anything. When an AI produces a polished answer, the ease of reading it feels like understanding. Users take that ease as a sign of their own competence, even though they didn't produce the work Does processing ease mislead users about their own competence?. An effort reward doesn't fix that illusion, because students already believe they're doing fine. Metacognitive feedback goes after the misperception itself. This connects to a wider pattern. A review of 14 studies found that students who lean too heavily on AI dialogue systems gradually swap careful reasoning for quick mental shortcuts Does relying on AI dialogue systems harm student critical thinking?. The harm builds quietly, which is why making it visible matters.

The corpus also points to a related approach: change what the AI says instead of what the student is told. Thinking assistants that paired advice with reflection questions beat assistants that only advised, only asked questions, or did neither Do reflection questions help people make better decisions with AI?. Similarly, systems built to point people toward the useful parts of a problem, rather than hand over a decision, avoided anchoring bias and kept the human doing the judging Can AI guidance reduce anchoring bias better than AI decisions?. Together these suggest a design principle. The fix isn't less AI. It's AI that keeps the learner's own reasoning involved.

There's a cost to watch for. Interventions that break into someone's thinking, even helpful ones, can disrupt their focus and force them to rebuild concentration Does AI assistance always help reasoning or does it carry hidden costs?. A metacognitive prompt that pops up at the wrong moment could create the same disruption it's trying to prevent. The corpus doesn't yet answer how long these effects last, or whether students who are told about the costs of offloading keep changing their behavior once the feedback stops. That gap is the most important open question here.