This is the best piece of its type that I've read so far. Whatever our emotional response might be to LLMs' effects on schools, it's what we do next that really matters.
Something I've been thinking about lately is to what extent AI can assist schools with developing their own digital systems for tutoring, feedback, record-keeping, etc. Should schools adapt their curricula or workflows to more closely match what an EdTech vendor comes up with or should we use a similar amount of energy to direct an agent to generate a bespoke system? The latter hasn't been realistic in many places before now. That doesn't necessarily make it better, but it does make things interesting.
I've been thinking more and more just how much guidance students really need when they use AI to actually make it worthwhile. On the one hand, many students are already power users but, in my observational experience, really don't know very much about the tech or why it does what it does. It's such a powerful tool when wielded in certain ways that it's kind of amazing to me that it was just placed in their hands a few years ago and so few educators responded to the new paradigm. Now that it seems like schools are ramping up, I share many of the concerns you identify in this piece - "plastic" policies that do very little to give teachers granular advice and so few who really know what they're doing. It will be very interesting to see what develops.
Thanks, Stephen. Yes I think you’ve nailed it. Students can look like “power users,” but often without much real understanding of what the tech is doing or where it falls short. That’s where the guidance matters most. Also my hunch is that any policies created in 2025 will be very quickly outdated by 2026.
This is the best piece of its type that I've read so far. Whatever our emotional response might be to LLMs' effects on schools, it's what we do next that really matters.
Something I've been thinking about lately is to what extent AI can assist schools with developing their own digital systems for tutoring, feedback, record-keeping, etc. Should schools adapt their curricula or workflows to more closely match what an EdTech vendor comes up with or should we use a similar amount of energy to direct an agent to generate a bespoke system? The latter hasn't been realistic in many places before now. That doesn't necessarily make it better, but it does make things interesting.
Thanks James. It certainly is an interesting time and I think things are moving rapidly.
I've been thinking more and more just how much guidance students really need when they use AI to actually make it worthwhile. On the one hand, many students are already power users but, in my observational experience, really don't know very much about the tech or why it does what it does. It's such a powerful tool when wielded in certain ways that it's kind of amazing to me that it was just placed in their hands a few years ago and so few educators responded to the new paradigm. Now that it seems like schools are ramping up, I share many of the concerns you identify in this piece - "plastic" policies that do very little to give teachers granular advice and so few who really know what they're doing. It will be very interesting to see what develops.
Thanks, Stephen. Yes I think you’ve nailed it. Students can look like “power users,” but often without much real understanding of what the tech is doing or where it falls short. That’s where the guidance matters most. Also my hunch is that any policies created in 2025 will be very quickly outdated by 2026.
"The machine is disgusting and we should break it."
https://anthonymoser.github.io/writing/ai/haterdom/2025/08/26/i-am-an-ai-hater.html