Two Years On, What Have We Learned about AI in Schools?
In short:
AI is here to stay but there’s no consensus on how to live with it (yet)
Early adopters confirm that thoughtful policies, staff upskilling, and honest dialogue about risks and rewards drive lasting impact
Enthusiasm arising from potential productivity boosts needs to be matched by caution about unseen effects on mental health and actual retained knowledge
ChatGPT just turned two at the end of 2024, and we’re seeing the first wave of data about how AI is reshaping work and learning. There’s a lot to be excited about, but news about teen mental health risks, AI-fuelled persuasion/radicalisation, and the risk of reliance on AI remind us to take a measured approach before rolling out untested tools and pedagogies beyond the pilot stage.
In this edition, we focus on three top-of-mind themes that emerged from discussions with educators in Hong Kong and Singapore, in the hopes of attracting broader perspectives from around the world.
Your thoughts are welcome at info@goodfuture.foundation.
1. Plagiarism & Academic Honesty
“My student’s grammar suddenly became flawless and their ideas actually make sense now, of course it’s AI.”
“I can’t assign take-home work anymore, we’re back to in-class assessments and exams.”
“I watched a YouTube tutorial about giving the AI my old homework so it can write like me.”
Perspective: It seems obvious in hindsight that Large Language Models would impact fields that rely heavily on language, like education. It’s also obvious that current norms around assessment will have to change, especially at the secondary and tertiary levels.
Teachers must know that AI writing detection does not work, and many universities have disabled the detection function in Turnitin (e.g., Yale, MIT, Berkeley, Vanderbilt, SMU, etc.).
The prognosis doesn’t have to be grim, teachers will find better ways to assess student performance—consider how calculators changed the way Maths is taught.
2. Staff Burnout
“The whole staff room is exhausted and if AI can save them a few hours a week, that’s great.”
“All the time I spent writing reports should’ve been spent on students.”
“I know I could’ve done more, but who has the time?”
Perspective: The good news is that AI productivity gains are real. Microsoft Research found that knowledge workers can save at least 30 mins a day using AI, and AI users also produce 10% more documents than before while spending up to 25% less time on emails (yes please!). Anyone who has tried any state-of-the-art model will know how AIs can blitz through lesson materials and reports; tools and workflows just need to be discovered first.
Because schools generate so much paperwork, the hardest part of implementing AI (documenting processes, keeping reports and minutes, digitising homework, etc.) is already half done. The challenge currently lies in the last mile, where tools don’t yet exist to help schools tie together all their information while making that accessible in a user-friendly way, but I have no doubt that learning management systems companies are all working on that.
3. Supporting Students
“Asking AI a question is less stressful than sending an email to my teacher. And I can get answers at 3AM.” (Ed: even though this is possible, we should prioritise sleep health and 'switching off' above all else)
“Their homework is better with AI, but how do I know if it reflects their comprehension or the AI’s?”
“How can we let students use all these AI tools when even the teachers don’t know anything about them?”
Perspective: All signs point to new and better forms of student support. A working paper from Nigeria shows that students who had never used a computer before achieved 2 years’ worth of learning gains in 6 weeks, by going to a computer lab twice a week to use GPT-4 as a tutor. However, a Wharton paper also showed that if you take the AI away, students can actually perform worse than those who didn’t use AI in the first place.
Regardless, earlier findings about how AI can help some entrepreneurs make more money, researchers make more discoveries, artists create more popular art will probably be echoed in education, as future research reveals all the ways that AI can make teachers and students more effective.
What’s Next?
If the history of the Internet and smartphones taught us anything, it’s that AI will remain a buzzword for about a decade. At some point, shiny new technologies will be taken for granted and we will just think, “of course I can generate 30 sets of assessments and worksheets and notes tailored to each student’s needs and interests. I just need to hit enter!”
 | A guest post by
|