Stanford urges teacher-first AI use

Stanford University education researchers are urging schools to keep teachers at the center of classroom AI use as districts and instructors sort through where the tools help and where they can undercut learning. A Stanford Report article published May 13 said researchers across the university are focusing on “evidence-based applications” that support “high-quality learning” and help educators do their work, rather than treating AI as a stand-in for instruction. Education Week, in a separate May opinion piece by veteran teacher Larry Ferlazzo, made a similar case for English learners, saying AI can help when its use is responsible and bounded. Together, the pieces outline a teacher-first approach: narrow tasks, close supervision, and human review of what the software produces. ### Why are Stanford researchers warning against open-ended AI use? Stanford’s May 13 article said AI in education is “rapidly evolving” and described a research agenda centered on risks as well as opportunities. The piece said experts are working with educators to understand how to use the technology thoughtfully, with an emphasis on meaningful learning and AI literacy. (news.stanford.edu) Mehran Sahami, chair of Stanford’s computer science department, said the tools add productivity and that students will use them even if schools try to ban them. But Sahami framed the question as instructional design, asking what form of teaching can build a solid understanding of the material while also taking advantage of the tools. That formulation leaves the teacher, not the chatbot, in charge of what students are supposed to learn. (news.stanford.edu) Karin Forssell, who directs Stanford’s AI Tinkery and GSE Makery, said schools are still in a period of “trial, error, and rigorous research” as they work out AI’s role. Her comments, as presented by Stanford, described a field still deciding what students should learn to do in an AI-rich environment. ### What does “teacher-first” look like in practice? (news.stanford.edu) Education Week’s technology section described Ferlazzo’s May opinion essay with a simple condition: AI can be “a great ally” for English learners “when used responsibly.” That phrasing points to the main guardrail in both accounts — the tool is useful when a teacher defines the task and the limits. Stanford’s article pointed to classroom uses that are tied to explicit learning goals, not unrestricted conversation. (news.stanford.edu) One example was Stanford’s introductory coding course, where students build an “Infinite Story” assignment and then use ChatGPT to extend it. In that case, the AI interaction comes after students have already done the core design and coding work. (edweek.org) The emerging pattern is narrower than the early pitch for AI as an all-purpose tutor. In these accounts, teachers decide the sequence, define the prompt, and judge whether the output is accurate or educationally useful. That conclusion is an inference from the way Stanford researchers and Ferlazzo describe acceptable use. (news.stanford.edu) ### Why does this matter especially for English learners? Larry Ferlazzo’s essay, as summarized by Education Week, focused on English learners, a group for whom practice, feedback and language scaffolds can be useful if they are well designed. The key condition in the summary was not the tool itself but the way it is deployed: responsibly. Stanford’s broader discussion of AI literacy also bears on that question. (news.stanford.edu) If students are using AI to generate language, code or explanations, teachers have to help them understand what the system is doing, where it can mislead them, and how to check the result. That is especially relevant in language-learning settings, where fluent-looking output can mask errors or bypass the productive struggle that teachers are trying to cultivate. (edweek.org) The point about checking and oversight is an inference drawn from Stanford’s emphasis on evidence-based use and meaningful learning. ### Is Stanford proposing more research, not just classroom rules? Stanford’s article said the university is pairing its public guidance with new funding. The piece said AI Meets Education at Stanford, or AIMES, and the Stanford Accelerator for Learning will offer $1 million in seed grants to faculty, students and staff to rethink how AI fits into college teaching. The grants will support course development, research and scholarly work on AI and education. (news.stanford.edu) That next step puts a date and a mechanism behind the debate. As Stanford researchers and classroom educators continue testing what works, the public guidance is available in Stanford Report’s May 13 article and Education Week’s May 2026 opinion coverage by Larry Ferlazzo. (news.stanford.edu)