After the noise of the school day ebbs, the work of grading begins. Afternoon coffee in hand, the teacher boots up, clicks into each file and gets a familiar, sinking feeling reading paper after paper of polished AI prose. Closing the laptop a little too hard, the teacher heads home for the day.
Cheating offends us. But concern about fairness and integrity masks a deeper threat.
AI in schools, without guardrails, erodes thinking.
Here’s what did not happen when students chose the AI shortcut: brainstorming, ideation, claim-making, evidence-gathering, word choice, sentence crafting and revision. All the cognitive work that writing develops. Apply that to areas beyond writing, and you skip the purpose of school.
And this shortcutting is happening now, with about 50% of students using some form of AI in their schooling. And according to a recent Rand survey, 80% of students say that teachers don’t explicitly teach them how to use it at all.
If students have AI, but no one’s teaching them how to think with it or putting limits on its role, what are they actually learning?
Human brains need friction to grow. When friction goes away, learning stops.
Offloading vs. outsourcing
This phenomenon has happened before. A new paper from Dr. Dirk Van Damme at the Center for Curriculum Redesign traces this pattern across other innovations, from writing to the printing press to calculators to GPS, and finds the same dynamic. Each technology expands what humans can do while simultaneously weakening the internal capacities that are no longer active.
AI bypasses foundational capacities like attention, memory and reasoning. These are non-delegable because learners need to develop them for understanding, judgment, freedom and autonomy.
And there’s a difference between cognitive offloading (where you still think and the tool supports you) and cognitive outsourcing (where the system thinks and you consume the result).
If students consistently bypass cognitive work, schools risk undermining the very capacities they are meant to develop.
The good news is that the same technology that erodes thinking can strengthen it, but only when it’s designed for learning and guided by humans who protect the cognitive friction. Before deploying any AI tool in any activity, educators should ask one question: Does this increase or decrease the cognitive work my students need to do? If AI is doing the hardest part, students aren’t practicing the hardest part.
In a recent study published in Nature, Harvard researchers demonstrated that when an AI tutor was built using the same pedagogical best practices as expert classroom instruction, students learned more in less time and reported higher engagement than in traditional active-learning classes.
The key variable was not the use of AI, it was the intentional design behind it. The risk is in deploying technology without regard for whether students are doing the cognitive work.
We must ask: Is the student still doing the thinking?
Provide AI guardrails that maintain friction
To preserve the healthy cognitive stress that stimulates learning, educators can implement the following guardrails for AI use in schoolwork.
Guardrail 1: Require cognitive ownership
Students should attempt work before AI assists. Require students to outline their chain of reasoning before receiving AI support. When students reach for AI before attempting problems themselves, they miss the productive struggle that builds metacognitive skills. Simply providing AI “explanations” doesn’t automatically prevent dependency — students still need task design that requires reasoning.
Guardrail 2: Design for dialogue, not output
AI should question students rather than answer, forcing them to exercise their critical thinking skills, memory and creativity to tackle cognitive problems. Structure activities so AI is limited to feedback, hints or examples – not full answers. Require students to explain what they changed and why.
Guardrail 3: Make thinking visible
Require students to show their work and explain their choices, down to individual words. Build in AI-free checkpoints to confirm independent understanding. Teachers should maintain real-time visibility into how students engage with AI, so they can identify where students struggle and whether understanding is genuine.
Guardrail 4: Teach AI literacy explicitly
Teachers should model how to prompt AI platforms in ways that promote learning and how to critique the outputs. Establish classroom norms requiring students to document their creative process and explain how AI tools contributed to their work.
Well-designed technology prompts learners to draft their own arguments first. Then raises tactful questions to help the student identify gaps in their thinking or suggests alternatives to try. It prompts them in dialogue to consider opposing points of view and to strengthen writing through revision.
Not only is friction preserved, but it is strengthened.
We are already seeing this happening in classrooms where teachers have full control of designing their AI interactions to ensure students own cognitive work. In a recent classroom study of over 23,000 teacher-created AI learning activities, 73% of lessons required conceptual understanding, while 59% asked students to analyze information and 58% asked them to evaluate ideas or make judgments. When teachers lead the design, AI can strengthen critical thinking, increase engagement and support responsible instruction across classrooms.
We are at an inflection point where students are able to bypass all cognitive work our schools demand if they choose. If schools become places where AI does the thinking, what are schools for?
I am optimistic because if we listen carefully to young people, we realize that this generation is good. They don’t want to cheat — they want guidance on how to use AI for thinking and learning, not outsourcing.
To realize the promises of AI in schools, we educators must protect friction that makes learning happen.
Opinions expressed by SmartBrief contributors are their own.
Subscribe to SmartBrief’s FREE email newsletter to see the latest hot topics on edtech. It’s among SmartBrief’s more than 250 industry-focused newsletters.