Jargon-ridden and light on genuine reflection, NYC Department of Education’s initial guidance on the use of artificial intelligence in the public schools augers a new frontier for strategic failure in a system long plagued by meager vision and reactive leadership.
Mar. 26, 2026 – The New York City Department of Education (DOE) yesterday published its initial guidance on artificial intelligence, a milestone which it describes as performing the Phase 1 work of setting “the foundations and boundaries.” In a mere three months, DOE promises, a “Phase 4” will see the publication of a full “playbook,” including guidance on the currently “under development” topics of “homework and academic integrity” and “cognitive offloading.”
The initial guidance is available online in two versions: one, visually appealing apparently designed for easy digestion; the other, full text.
My review of the initial guidance and the path it portends: Staggeringly lacking. (I would say “comically bad,” but for the fact that the stakes for hundreds of thousands of students now in the public-school system, and for millions of students to come, are so high.)
Fatalism and inevitability framed as realism
We’ve seen this movie before. As with the argument that globalization was inevitable, and we just needed to get on board. Or that the retirement age for Social Security needs to be raised rather than lifting the cap on wage income subject to FICA. Or that having affordable healthcare insurance and a primary care doctor is an extravagance we can no longer tolerate, so we should accept limited insurance networks and welcome a revolving crew of (mostly non-M.D.) “providers.”
Here, DOE puts it just as starkly. “The question is not whether AI belongs in schools,” we are told. Instead, “The question is whether we will collectively build a system that governs AI to serve every student and every stakeholder.” (See sidebar on DOE jargon.) Or, as it is stated in the full-text version: “Our students are already encountering AI beyond school walls. The question is whether they are equipped with critical thinking, ethical grounding, and creative agency, or whether they are left to navigate AI alone.”
But, of course, the foundational question is precisely whether AI belongs in schools (and, if so, at what grade level and to what extent); presenting the choice as “accept it” or leave students to “navigate it alone” is a false binary.
Cart before the horse
Having already decided that AI for students is the way to go – and throwing in some AI boosterism for good measure: there’s a commitment that the system “builds AI literacy” for students – threshold questions are ignored.
What if DOE’s deeply inadequate process of helping students to learn and thrive, to help each reach his or her potential, and to nurture them to become college- or job-ready, as well as to become active and informed citizens, needs far more basic remediation than AI can patch? What if AI – not just in elementary school, but in middle school as well – is little more than a bright, shiny bauble that pretends to offer solutions but ignores the basics?
AI is not going to help teachers and administrators learn that reading skills can’t be built on a foundation free of subject-matter knowledge. AI is not going to help teachers demand more from their students or help students have the stamina and patience to do more than read short passages.
AI is not going to adapt school schedules so that middle schoolers have significantly more time for physical activity.
AI won’t deal with the unwillingness in the system and in the political class to remove teachers who are mediocre and worse from the classroom (and the payroll).
And to the extent AI is deployed to help the DOE continue to pretend that you can have students of vastly different levels of readiness thrown together in a room with the result that all actually learn to their potential, we would see exactly the opposite of promises of humans remaining central: having each kid, or grouping of kids, alone with an AI “tutor” is not the promised land.
Radical reimagining needed
It is true that the initial guidance offers a glimpse into the broad subject areas of work in progress (although nothing about even the most preliminary practical frameworks). Thus, the commentary on “cognitive offloading” states that, “Effective AI integration preserves the intellectual work of learning. NYCPS is developing research-informed guidance on instructional design that ensures AI supports – rather than substitute for – student thinking.”
A blurb on “homework and academic integrity” states that, “NYCPS is developing guidance on assessment design and academic honesty in an AI-enabled environment, grounded in existing academic policies and research on effective practice.” (What grade would you assign that sentence if a student or employee submitted it to you?)
I, for one, am not reassured. We’re supposed to get from here to a concrete “responsible use and guide rails” policy by June?
One might have thought that DOE would have learned something about the perils of chasing tech after a generation of deeply pernicious effects of screens and phones.
But, at best, DOE remains surprisingly naïve about how difficult it is to keep AI within any reasonable bounds. Does it believe that teachers will reliably understand that AI models are currently designed to cater to what the user wants to hear, and resist the sycophancy of ChatGPT or Gemini or whatever other AI iteration comes down the block? 10-year-olds? 14-year-olds? 18-year-olds?
How can a student who has never learned what it is to ask and ponder questions on his own either learn to do so when AI is beckoning or learn to be critical of AI output?
The issue, and it is not limited to DOE or to K-12 education by any means, is one of fundamental philosophy. The educational establishment argues at the same time that there is nothing it can do to change the fact that AI is in the world for students to access, and that some tweaks to regulations will provide adequate guide rails for school purposes.
Entirely lost is the idea that schools have any role in insulating students from the outside world. That the things you do to begin to become a full-fledged human may (ought to) differ from the range of things you do once you’re much further down the road of having become a full-fledged human. That the things you do to prepare to be in the world may (ought to) differ from the range of things you do one you are fully in the world.
That insulating function is easier to implement in elementary school (“just say no”) than thereafter, but principles are simpler at all levels. Unless and until we have a society closer to one that frowns on cheating and honors integrity (and we seem to be move rapidly in the opposite direction), more of what is traditionally considered homework (including research projects at upper levels) will need to be done in school, and that means wholesale revamping of school schedules (45-minute class blocks can’t work for teachers or students in an AI-suffused world). Even more teacher time is going to have to be spent – again, at all levels – to determine quickly which students are doing the reading and which students are having AI do the reading for them.
Even if there were any willingness to embark on this project – and, quite evidently, there isn’t – it certainly wouldn’t be complete by June. As such, we’ll have another area of pretend: AI is being controlled, just like graduation rates are real.
How interested is DOE in high-quality instruction anyway?
The full text of the guidance does bow in the direction of the need for “rigorous” learning and instruction in several places, but the relationship of this goal and AI tools is entirely unclear.
Moreover, when the guidance describes six categories of risks of AI for students, “erosion of the thinking, creativity, and problem-solving skills students must develop themselves” is mentioned last of the six.
Neither “instructional effectiveness” nor “academic excellence” has its own category in the list of questions DOE is “still working to answer,” with the former subsumed (as the last-mentioned item) is the category called “bias and equity review.” The closest any of the in-progress items comes to academic excellence is the reference in the cognitive offloading section to the claim that DOE is “developing research-informed guidance on instructional design that ensures AI supports, rather than substitutes for, student thinking.”
As for the four core commitments that the full-text guidance advert to, one sets out “Foundational literacy and math, meaningful learning experiences, social connection, and preparation for life beyond school are the building blocks of student opportunity.” Each of the elements of this commitment — currently honored in the breach for vast numbers of students — is important, but it is very different from a system that would say that the standard DOE builds to in every decision is maximizing the academic development of each student to his or her full potential.
It is easy to see a few years down the road to a time when the underlying inadequacies of the system to excel in meeting the needs of low-readiness, medium-readiness, and high-readiness students will remain as they are today, with AI implementation providing a convenient distraction to tell the look-how-cutting-edge we are story.
A key peril of going down the AI path
An educational philosophy or program, once in motion, tends to stay in motion. We saw that with “balanced literary.” Long after it was clear that the approach did not work for large numbers of students, DOE, like many other school systems, continued to use balanced literacy. It was only when opposition to the status quo managed to gain a critical mass — perhaps 20 years after implementation — that the approach was abandoned.
A status quo acquires an entire ecosystem of reinforcers. AI vendors are prominent among them, but the forces of self-justification cannot be minimized. Once an institution has bought into AI for students — as DOE has — the temptation is enormous to oversell successes, underplay problems, and tenaciously refuse to reconsider first principles. There follows a long period where various adjustments are tried, but where, students and parents are told, there is no going back.
We now have a generation that finds it hard to maintain attention, hard to read a whole book, hard even to socialize face-to-face. It is not difficult to imagine how the thinking, creativity, and problem-solving skills students must develop themselves” will erode if DOE does not take a very significant time out.