What the research says about AI and student thinking

Ask whether AI helps or hurts student learning and you start an argument. Ask what the evidence is actually beginning to show, and you get something more useful: a specific worry, a named mechanism, and a clear line between the uses that build thinking and the ones that quietly erode it.

This page summarizes what recent research and large educator surveys are finding, defines the mechanism most often blamed, and connects it to the concept this program exists to name: judgment displacement. The research is early and much of it is correlational, so we have tried to state it carefully rather than for effect.

The concern is now near-universal among educators

The clearest signal so far comes from how widely students use these tools and how consistently their teachers worry about it. In an October 2025 College Board brief, 84% of high school students reported using generative AI for schoolwork such as brainstorming, revising, and research. A February 2026 College Board survey of more than 3,000 college faculty found that 74% report students using AI to write essays or papers, and that concern is close to unanimous: more than 84% of faculty agreed AI reduces students' critical thinking, originality, and deep engagement with the material, 88% were concerned about overreliance on automation, and 92% were concerned about AI-assisted plagiarism.

Educator concern is not the same as proven harm. But when the people who watch students think for a living report the same pattern at that scale, it is worth taking as a serious signal rather than a panic.

Cognitive offloading: the mechanism behind the worry

The behavior underneath the concern already has a name in cognitive science: cognitive offloading, the act of handing a mental task to an external aid instead of doing it in your head. Writing down a number instead of memorizing it is cognitive offloading. So is letting a calculator hold the arithmetic, or letting a map hold the route. It is often useful, and it is not new.

What is new is a tool that will offload not just memory or arithmetic but the reasoning itself. A 2025 study published in the journal Societies reported that more frequent use of AI tools was associated with weaker critical-thinking performance, with cognitive offloading appearing to mediate the relationship. Correlation is not proof, and heavy AI users may differ in other ways. But it lines up with what educators are reporting and with a simple intuition: a faculty you stop exercising is a faculty that weakens.

From cognitive offloading to judgment displacement

Cognitive offloading names the behavior. It does not, by itself, tell you when the behavior costs something. Offloading a route to a map costs you little. Offloading the deciding, the part where you form a view of what is true and what you will put your name on, costs you the very thing school is supposed to build.

That is the distinction this program draws. When offloading lands on the deciding, we call it judgment displacement: the tool stops assisting a decision and starts making it, and the student produces work without ever forming a view of their own. Repeated across months and years, those moments add up to a trend we call cognitive drift, the slow, mostly unnoticed loss of the habits that let a person think for themselves. The research measures the behavior; these terms name what it costs and where.

Why students are the exposed case

An adult who offloads reasoning to AI is leaning on judgment they already built. A student is offloading judgment they have not built yet. Adolescence and early adulthood are exactly when these habits form, and they form in the moments AI is best at taking over: the essay that needs a real view, the problem with no answer key, the decision about what to trust. Hand those moments to the tool and you are not shedding a habit, you are skipping the years the habit was supposed to form in. That is why the same tool that merely dents an expert can quietly reshape a student.

What the evidence does and does not say

It does not say AI is the enemy of learning. Used well, it is a genuine aid: a starting point, a checker, a way to generate, verify, revise, and decide. That loop keeps the student in charge, and the research does not condemn it. What the evidence points to is narrower and more useful. The variable that matters is not how much a student uses AI. It is where the tool sits when they do, and whether they keep the checking and the deciding or hand them over.

The response

If the risk lives in the deciding, the response has to protect the deciding. That is what the five principles are for, and why schools distribute them: think with the tool, not through it; protect what is private; verify before you trust; follow the rules of the room; and know when to stop and ask. They are small enough to carry into the moment that matters, which is the only place the research suggests it is decided. Schools can bring the program in at no cost and with no student data collected.

Sources and further reading

College Board, faculty perceptions of generative AI in higher education (February 2026) · College Board, high school students and generative AI (October 2025) · Societies, AI tools and cognitive offloading (2025) · UTS, unstructured AI use and cognitive atrophy (2026).