AI Literacy Before AI Dependence

Why I support Mayor Zohran Mamdani’s decision to slow the use of generative AI in New York City schools

The more I collaborate with artificial intelligence, the more convinced I become that schools need to be careful about where it sits in the learning process.

That may sound contradictory. It isn’t.

I use AI regularly. I ask it to challenge an idea, identify a blind spot, suggest a different structure, simulate a sceptical reader or open possibilities I had not considered. Used this way, it can be a remarkable brainstorming collaborator. It sharpens thinking when thinking is already underway.

But I have also seen how easily it can step over that line.

AI can move from supporting a thought to supplying one. It can move from helping a student develop a voice to producing a polished imitation of one. The final answer may be competent—even impressive—while the learning underneath it remains dangerously thin.

That is why I support the direction taken by New York City Mayor Zohran Mamdani and Schools Chancellor Kamar Samuels.

On 2 September 2026, New York City announced a one-year moratorium on student-facing generative AI for children from 2-K through Grade 8. High school students will instead receive AI-literacy instruction, while a limited number will participate in carefully supervised pilots using vetted educational tools. The policy is not a permanent ban. It is a deliberate pause designed to protect younger students while the city studies how AI affects learning, development and classroom relationships. New York City Government

I believe that is a thoughtful and educationally responsible position.

A pause is not a rejection of technology

Some people will inevitably interpret the moratorium as anti-technology.

I see it as something much more intelligent: educational sequencing.

Before students begin delegating parts of their thinking to a machine, they need to understand what thinking feels like.

They need to experience the frustration of not immediately knowing. They need to search for language, sustain attention, make a mistake, return to a problem, listen to another person, revise an idea and discover that their first answer was not necessarily their best one.

These experiences can feel inefficient. They are also where intellectual independence begins.

A child who has not yet developed judgement cannot meaningfully exercise judgement over an AI-generated response. A student who has not formed a voice will struggle to recognise when a machine is quietly replacing it.

There is nothing backward-looking about protecting the developmental stages through which human judgement, language, creativity and confidence are formed.

The greater danger is not simply cheating

Much of the educational conversation about AI has focused on plagiarism and cheating. Those are real concerns, but I do not believe they are the deepest ones.

The more significant risk is premature closure.

Generative AI offers immediate coherence. It can produce a fluent answer before a student has properly formed the question. It can close down uncertainty before curiosity has had time to deepen. It can transform the blank page from a space of possibility into an inconvenience to be eliminated.

In creative education, the blank page is not a failure state.

It is where ownership begins.

The uncertain opening of a project—the searching, sketching, discarding, discussing and trying again—is not wasted time. It is the work. When AI removes that process too early, it does not simply make the task easier. It may remove the very experience the task was designed to teach.

We should not confuse the production of a polished outcome with the development of a capable mind.

AI should remain a collaborator

My own position has become clearer through continued use: AI is most valuable as a brainstorming collaborator, not as the author, designer or substitute for a student’s intellectual labour.

That boundary is not anti-AI. It is what makes AI educationally useful.

In a design classroom, for example, students might use AI to generate research questions, identify possible audiences, test assumptions, locate gaps in an argument or receive an alternative critique of a developing concept.

But the student must remain responsible for the consequential decisions.

They must select, reject, justify, sketch, prototype, photograph, write, discuss, revise and make. They should be able to explain where AI was used, what it contributed, what they disagreed with and how their own thinking changed.

The evidence of thought must remain visible.

At its best, AI widens the field of possibility. It should not quietly make the journey on the student’s behalf.

AI literacy must come before AI dependence

The high-school component of the New York City policy is particularly important. Rather than pretending that teenagers will not encounter AI, the city intends to introduce twice-yearly literacy modules and limited teacher-supervised pilots. New York City Government

That is a sensible beginning. But it must be only a beginning.

AI literacy cannot be reduced to teaching students how to write an effective prompt. Prompting is a technical skill. Literacy is much larger.

Students need to understand that generative AI predicts plausible language rather than possessing human knowledge or judgement. They need to recognise confident errors, fabricated information and hidden assumptions. They need to examine how training data can reproduce cultural bias and whose knowledge may be absent, distorted or appropriated.

They need to understand privacy, data ownership, authorship and disclosure. They should learn how to verify claims, trace genuine sources, identify synthetic media and question the persuasive fluency of an answer that merely sounds authoritative.

They also need opportunities to reflect on dependence.

What happens to my thinking when I use AI immediately? What happens when I wait? Has the tool helped me develop my idea, or has it relieved me of developing one? Can I defend this work without returning to the chatbot? Does the final piece still sound like me?

No student should leave school merely knowing how to operate AI.

They should know how to question it, constrain it, verify it, disagree with it and decide not to use it.

That is genuine AI literacy.

Teachers need this literacy as well

Schools cannot ask students to use AI responsibly while leaving teachers without the time, knowledge or shared language needed to guide them.

Extensive professional learning will be essential. Teachers need to understand what these systems can do, where they fail, how student data may be handled and how AI changes the design of assessment.

We also need to move beyond a culture of trying to catch students using AI.

The more productive question is: How do we design learning so that thinking remains visible?

That may mean more discussion, oral defence, process journals, annotated drafts, classroom observation, peer critique, practical production and spontaneous questioning. It may mean assessing how a student arrived at an outcome rather than concentrating almost entirely on the polish of the final submission.

Authenticity cannot simply be policed into existence. It needs to be designed into the learning process.

This is particularly important in subjects such as design, where iteration, decision-making and reflection are not supporting evidence around the task. They are central parts of the task itself.

Keeping education human

Schools do not exist to make every intellectual task frictionless.

They are among the few remaining places where young people can practise sustained thought, conversation, patience, disagreement, failure and recovery in the presence of other human beings.

AI has a place within that environment. But it should not sit at its centre.

I support Mayor Mamdani’s decision because it refuses the assumption that technological adoption is automatically educational progress. It gives younger students space to develop foundational capabilities before introducing a tool that can so easily imitate them. For older students, it proposes supervised access alongside critical education rather than unrestricted dependence.

The more I work with AI, the more I value what it cannot replace: the original question, the uncertain mind, the conversation with a peer, the attentive teacher and the slow formation of independent judgement.

AI should be something students learn about before they are routinely asked to learn through it.

And when it does enter the classroom, it should expand human thinking—not quietly take its place.