In January 2023, New York City's Department of Education banned ChatGPT from school devices and networks, citing cheating and concerns about critical-thinking skills. By May of the same year, the district reversed the policy, with schools chancellor David Banks writing publicly that the initial ban had been an emotional reaction that missed the potential of a tool students would need to understand regardless of school policy.
That four-month reversal is a compact case study in how public conversation about new technology keeps collapsing into two postures, because each is easier to hold than the alternative. Optimism without qualification lets you stop asking what could go wrong. Panic without qualification lets you stop asking what could go right. NYC's initial ban was the panic version; its reversal was an attempt, imperfect, to walk back toward the harder middle.
The actual record of new capabilities rarely sorts cleanly into either camp. Most consequential technologies have produced real gains for some people, real costs for others, and a long middle period where it was genuinely unclear which effects would dominate — a period during which loud certainty on either side did not actually help a single classroom navigate the transition.
What gets lost when discourse collapses into these two cartoons is the harder, more useful work: naming specific mechanisms, specific groups affected, and specific tradeoffs, rather than rendering a verdict on the technology as a category. A tool that displaces one kind of homework while opening a different kind of literacy is not well described by either 'ban it' or 'embrace it.'
Holding a position of genuine hope without denial requires tolerating open questions for longer than either cartoon allows — believing something is likely to be net beneficial for how students learn while still tracking, in detail, which students are exposed to which costs, and whether anyone is actually addressing them.
This is harder to write into a single policy memo, harder to headline, and harder to defend at a school board meeting, which is precisely why so many institutions default to the simpler postures instead — ban first, or adopt uncritically, rather than build the more tedious middle position.
NYC's district did not land on a fully resolved answer. What it landed on, eventually, was a curriculum that teaches AI literacy directly rather than pretending students will not encounter these tools, alongside continued and specific rules about when their use in an assignment counts as cheating.
That is not a triumphant ending, and it should not be described as one. It is what is left after an institution stopped reaching for the nearest available verdict and started asking, tool by tool and assignment by assignment, what the actual tradeoff was — which took four months longer than the ban did, and produced something that has actually lasted since.