Every teacher now knows the genre. A student who struggles to write a coherent paragraph in class hands in homework that reads like a briefing paper from a management consultancy: flawless grammar, five-paragraph structure, and a vocabulary the child has never once used out loud. AI has made school easier for students, and in doing so it has broken something old and important — the relationship between effort and learning.
The cognitive science here is not new. Robert Bjork spent decades demonstrating what he called "desirable difficulties": conditions that slow learning down in the short term but produce deeper, more durable knowledge in the long term (Bjork, 1994). Retrieval, generation, effortful reconstruction — the struggle is not an obstacle to learning, it is the mechanism. The converse is also true, and every teacher intuitively knows it: if you get it fast, it doesn't last. A student who receives a polished essay from a chatbot in thirty seconds has acquired a product and skipped the process. The product was never the point.
The case for the pen
The same logic applies to the humblest tool in the room. Mueller and Oppenheimer's well-known study (2014) found that students who took notes longhand performed better on conceptual questions than those who typed — because typing encourages verbatim transcription, while the slowness of handwriting forces you to select, compress and rephrase. To be fair-minded: later replications have complicated this picture. Morehead, Dunlosky and Rawson (2019) replicated the verbatim-transcription finding but not the performance advantage, and a large multi-lab replication (Urry et al., 2021) found no immediate benefit for handwritten notes at all. The honest reading is that the effect is smaller and more context-dependent than the headlines suggested.
Yet the underlying principle survives intact: shallow processing produces shallow learning, whichever device you hold. And on the narrower question of handwriting itself, recent neuroscience is suggestive — Van der Weel and Van der Meer (2024) found, using high-density EEG, that handwriting engages far more widespread brain connectivity than typewriting, in regions linked to memory and sensory integration. Forming letters by hand is slow, effortful and, apparently, exactly the sort of desirable difficulty Bjork had in mind.
The split I have landed on
Based on my experience so far, I have settled on a simple division of labour between the analogue and the digital. In the classroom: zero AI, zero student technology. All writing is handwritten, in books, in ink. Exams are handwritten too — not out of nostalgia, but because a handwritten exam is the only assessment format left that shows you exactly what a student actually knows, unassisted, under mild pressure. It is the last honest signal we have.
Homework is where the battle is actually fought. I time assignments so that a suspiciously fast return is itself a flag — a forty-minute essay submitted in twelve minutes tells its own story. Digital submissions go through tools with copy-and-paste disabled wherever possible. And I say plainly to students: if you want to use AI for research, fine — gather, compare, interrogate. But the organising of your thoughts, the structuring of your argument, the crafting of the sentences — that must be your own brain doing the work, because that work is the education. The writing is not the output of the thinking. It is the thinking.
None of this works without vigilance from teachers, and this is the uncomfortable part. AI offers us the same easy life it offers our students: let them generate, let it mark, and everyone goes home early. A profession that takes that deal will not notice its own hollowing-out until the hollowing is done. Our job, more than ever, is to be the adults who insist on the hard way — kindly, patiently, and without exception.
Pen and paper in class. AI, if they must, at home — for research, never for the writing itself. It is not a Luddite's charter. It is simply taking seriously the oldest finding in the science of learning: the brain keeps what it struggles for.