Clarendon hopes that you always give this missive the reverence it bloody well deserves by printing it out. However, given the subject of today’s newsletter, me must now insist that you do.
Screens are becoming more and more prevalent in our children’s education. From early literacy and numeracy apps, to Atom Learning and Century Tech for the 11+, GCSE and A level homework is regularly set on online platforms, and Clarendon is always struck whenever we visit school classrooms at how many screens are on display.
Let me be clear, Clarendon is not some crazed luddite. I haven’t written this with my quill and sent it via my carrier pigeon (Sir Humphrey) to the team in the office. Screens and AI have a crucial role to play in education, both today and in the coming years: delivery of content, cutting down marking time so teachers can teach, allowing students more autonomy on their educational pathway, allowing more students to access great teaching, and much more besides.
However, I believe as educators we must be extremely thoughtful and clear-sighted about when, how and why we use tech / AI. One only has to glance at the steady stream of (presumably AI created) dross on LinkedIn to witness the wild vanguard of educators calling for a complete overhaul of educational praxis to equip students with the tools for the coming AI revolution. It sounds nice, but in practice seem to always involve: tweaking prompts, debating an LLM and spotting hallucinations. I happen to think all of these are quite useful exercises in teaching AI literacy and illustrating the limitations of LLMs. However, there is a paradox here that I rarely see acknowledged: our students need good ol’ fashioned knowledge to effectively execute these tasks. And there is growing evidence to suggest that deep learning may be more effectively done away from AI and screens altogether.
- A recent MIT study (with limited numbers admittedly) showed that students who prepared and planned for essays physically (on paper) performed better than those using AI. EEG scans showed that neural activity was far more widespread and deeper for those not using AI, showing that the activation of neural pathways that cements things in our memory is not triggered as much by the superficiality of AI learning.
- A 2025 study of AITS (artificial intelligence tutoring systems) found that although students reported greater engagement when using an AITS, there was often a decrease in performance. There was increase in performance when the AITS were aided by human teaching.
- And there is a mountain of evidence that suggests we read better on paper than on screens. With participants exhibiting greater understanding and recall, especially with more complex text.
In conclusion, we absolutely must equip our children with the skills to navigate the AI world of their futures, but, paradoxically, this will not be achieved by getting them to use AI all the time and setting all their work on screens.