Some educators are cautiously experimenting with generative AI, questioning whether the tech can be used in classrooms without undermining numeracy and literacy
When Lauren started researching the British designer Yinka Ilori for a school project earlier this year, she was able to consult her new study pal: artificial intelligence. Tweets: @proferictalley and @mlhennessey56 Tweets: @proferictalley : Educators are definitely going to have to grapple *soon* with generative AI and what role and can / should / shouldn't play in educational pedagogy, assignment completion, and student evaluation. From today's FT: https://www.ft.com/... https://twitter.com/... Michael Hennessey / @mlhennessey56 : Great to see @VarsityUK's ChatGPT survey getting a mention in today's FT long read https://www.ft.com/...
Context & Ripple Effects
Classroom adoption has been running ahead of any settled pedagogy since early 2023, when a New York City public school for girls began having students critique ChatGPT-generated lesson plans and a Wharton professor made essay-writing with the tool mandatory so students would discover its accuracy and bias problems firsthand. The FT report lands mid-arc: educators like Lauren, researching Yinka Ilori with an AI study pal, are now testing whether supervised use can coexist with building real numeracy and literacy rather than substituting for it.
The institutional layer is filling in around these experiments. OpenAI has moved from product to training provider with its free online course for K-12 teachers, while US school systems weigh partnerships with teachers' unions as vendors push into classrooms — making the cautious-educator camp the swing constituency between vendor-led adoption and outright refusal.
First-order effects
- Teachers adopting the Wharton-style approach shift assessment design toward assignments that force students to verify and correct AI output, turning the chatbot's known accuracy and bias failures into teaching material rather than a cheating risk.
- Schools still without a policy face immediate pressure to choose between the supervised-use experiments and the rejectionist stance some academics have taken.
Second-order effects
- OpenAI's free teacher-training course and vendor investments in union partnerships become the counterweight: by supplying curriculum support directly, the company lowers the cost for skeptical districts to say yes, competing for legitimacy against educators who doubt the tech belongs in class at all.
- If supervised-use results show literacy or numeracy erosion, expect districts to swing back toward restriction, forcing vendors to reposition AI as a teacher-side tool rather than a student-facing one.
Third-order effects
- The pattern points toward 'AI verification' becoming an explicit taught skill — with the Wharton fact-checking model as the template — meaning curricula and standardized assessment eventually reorganize around evaluating machine output instead of banning it.
- A durable split is forming between institutions that integrate generative tools under supervision and communities that refuse them on ethical grounds, a divide already visible among creatives and academics weighing boycotts they may not sustain.
The trend: Education is moving from blanket bans toward supervised integration of generative AI, with pedagogy — not vendor capability — now the limiting factor on adoption.