9/12/2026
Around China's Teachers' Day, I came across an interesting piece of news from New York City, where I live.
Starting this school year, New York City Public Schools is placing a one-year moratorium on student-facing generative AI for students from 2-K through eighth grade. At the same time, the city is putting limits on individual screen time for younger students.
The policy takes a different approach with high school students. The city will require AI-literacy instruction and allow limited, supervised use of several approved AI programs. Every high school student will receive two 45-minute AI-literacy modules covering topics such as how AI works, bias, privacy, risks, and appropriate use.
Teachers can still use approved AI tools for lesson planning and operational tasks, but AI cannot be used for grading, behavior monitoring, or decisions about students' placement, promotion, or graduation.
The policy also prohibits AI companion chatbots across all grades and limits individual screen use for younger students. Students in grades 2-K through second grade will have no one-to-one screen time, while the city recommends no more than 30 minutes a day for grades 3–5 and 45 minutes for grades 6–8. There are exceptions for assistive technology and other necessary uses.
This made me think about a much more fundamental question:
What do children actually need from school?
If AI can answer almost any question a child asks, why do children still need teachers?
Because education is not simply the transfer of information.
A child can ask ChatGPT to explain fractions. AI may provide an accurate explanation in seconds. But an elementary-school teacher can notice that the child is frustrated. The teacher can encourage the child to try again, discover exactly where the confusion lies, adjust the explanation, and perhaps teach the child something that has nothing to do with fractions:
Don't give up just because you don't understand something yet.
AI can answer questions. A teacher can help a child become someone who knows how to find an answer, how to keep trying when the answer doesn't come easily, and how to learn from mistakes.
That distinction matters particularly for young children.
At the elementary-school age, children are still developing their confidence, curiosity, habits, and attitudes toward learning. They are learning not only what to know, but also how to think, how to ask questions, how to work with others, and how to deal with difficulty and frustration.
This is why people have always thought elementary-school teachers have a particularly important role. They are helping children form their first ideas about themselves as learners:
Am I good at learning? Is it safe to ask questions? What happens when I make a mistake? Can I solve something difficult? Do I give up when I don't know the answer?
The NYC policy explicitly puts some of these ideas at the center of its approach. For younger students, the schools say learning should emphasize human interaction, play, discussion, creativity, and the opportunity to work through difficult problems themselves rather than simply outsourcing those problems to a machine.
I find that particularly interesting.
AI may become one of the most powerful tools children will ever have. Before giving children powerful tools, we should make sure they have developed the ability to think for themselves.
They need to experience the frustration of not knowing the answer.
They need to struggle with a difficult problem instead of immediately asking a machine to solve it.
They need to learn how to work with another person, listen to a different point of view, make mistakes, and try again.
And they need to experience something AI cannot provide: a human being who knows them, notices them, encourages them, and believes in them.
The more important question is:
What should children learn to do for themselves before AI starts doing it for them?