AI Should Make Learning Messier, Not Easier

‍Guest blog by Adriana Perusin

Every new wave of educational technology comes with the same promise: learning will become smoother. Fewer obstacles. Faster feedback. Cleaner answers. With generative AI, that promise is even stronger. A student can ask a question and receive a fluent explanation in seconds. A teacher can produce a worksheet, rubric, or lesson sequence almost instantly.

That is useful. But it is not the same as learning.

The moments that change students are rarely smooth. They are often messy in the best sense of the word. 

The necessary friction 

Students argue over evidence. They test an idea and find that it does not work. They explain something badly, hear a better explanation from a peer, then try again. They get frustrated, but not abandoned. They revise. They notice their own thinking.

If AI removes all of that friction, it has made school more efficient and less educational.

My position is not that students should avoid AI. They are already using it, and pretending otherwise helps no one. Students should use AI when the task asks them to reason, critique, compare, iterate, or make a judgement. They should use it to test an argument, expose a weak assumption, generate a counterexample, or ask for another way into a difficult concept. The line is different: AI should not become an answer machine that lets students leave the thinking untouched.

Intentional design 

For teachers, the real opportunity is lesson design. During my years training teachers in active learning, I saw that the best lessons often looked untidy from the outside. A Socratic discussion can look noisy. Project-based learning can look slow. A gallery walk can look like movement without enough control. But good active learning is not chaos. It is designed uncertainty.

The teacher has made decisions in advance: what question is worth struggling with, what roles students will take, where misconceptions are likely to appear, how evidence will be shared, when the discussion should pause, and what students must produce by the end. That planning is demanding. It is one reason many teachers fall back on safer, cleaner formats even when they believe in deeper learning.

This is where AI can help. It can help teachers create better prompts, anticipate misconceptions, build scenarios, draft debate roles, vary examples, and design reflection questions. It can help students enter the work with more support and leave it with more ownership. Used well, AI does not tidy away the struggle. It makes the struggle more purposeful.

A human centre of gravity 

The future I want is not a classroom where every student sits silently with a chatbot. It is a classroom where AI has helped the teacher design richer human work: discussion, collaboration, critique, revision, making, performing, explaining. The technology may be present in the preparation, and sometimes in the activity itself, but the centre of gravity remains human.

We should judge AI in education by a simple question: does it make students think less, or think better?

If the answer is less, we should be sceptical, however impressive the tool looks. If the answer is better, the classroom may become louder, slower, and less polished. That is not a failure of AI. That is the point.

Key Takeaways

1. AI should not make learning frictionless. It should make productive struggle easier to design. And like in science, when learners have friction, it generates energy. 
2. Students need to make informed choices of when to use AI and when not to. They may be able to use AI well but when the task asks them to reason, critique, iterate, or make judgement calls, that likely needs to sit with them. 
3. Teachers need support designing structured uncertainty, not just faster worksheets. There is benefit in offsetting menial tasks but we probably should ask if the menial task is necessary! 
4. The best AI-enhanced lessons may look messy because students are debating, revising, building, and explaining.
5. The practical test is whether AI helps students think better, not simply finish faster.

AI is not the devil to be spurned. It is dynamite - in the right hands and conditions, it can be a force for good; in the wrong, it can be explosively devastating. We need to choose wisely not just for the future of education but also for the present. 

Adriana Perusin is co-founder of Flip Education and an education designer with more than 20 years of experience. She previously founded and led IASEA in Brazil, where she trained more than 1,000 teachers in active learning and socio-emotional learning. She writes about pedagogy, learning design, and the responsible use of AI in classrooms.