Cheat sheet

The most capable writing tool ever offered to schoolchildren has been withdrawn from them by the one institution that exists to teach them to write. On 3rd January New York City’s education department confirmed that ChatGPT, a chatbot from OpenAI, a San Francisco startup, is now blocked on its devices and networks.^{1Chalkbeat New YorkNYC schools ban ChatGPT, writing tool that uses artificial intelligence3 Jan 2023} Individual schools may ask for access if they wish to study the technology behind it.^{1Chalkbeat New YorkNYC schools ban ChatGPT, writing tool that uses artificial intelligence3 Jan 2023} Everyone else gets a blank page, which is where the cheating problem began.

The reasoning, offered by a department spokeswoman, Jenna Lyle, runs as follows. The tool “may be able to provide quick and easy answers to questions”, but it “does not build critical-thinking and problem-solving skills”, and there are “concerns regarding the safety and accuracy of content”.^{1Chalkbeat New YorkNYC schools ban ChatGPT, writing tool that uses artificial intelligence3 Jan 2023} The worry is not frivolous. Some teachers, according to the Washington Post as relayed by Vice, were in “a near-panic” about students using the bot to do their homework.^{2Vice (Motherboard)NYC Bans Students and Teachers from Using ChatGPT4 Jan 2023} We think the panic is understandable and the ban is wrong, for three reasons.

The first is that it does not work. The restriction covers department-owned devices and department networks.^{1Chalkbeat New YorkNYC schools ban ChatGPT, writing tool that uses artificial intelligence3 Jan 2023 2Vice (Motherboard)NYC Bans Students and Teachers from Using ChatGPT4 Jan 2023} It does not cover the phone in a pupil’s pocket, the laptop at home or the mobile signal outside the school gates. Any teenager will have noticed. A rule that binds only the honest and the lazy-but-law-abiding is a tax on compliance, and the department has levied it in the one place where a teacher can watch what a child does with the tool.

The second is that the department’s best argument cuts the other way. Accuracy is a real flaw. Stack Overflow, a forum where programmers swap answers, barred ChatGPT-written replies in December because its moderators judged that “the average rate of getting correct answers from ChatGPT is too low”.^{3Vice (Motherboard)Stack Overflow Bans ChatGPT For Constantly Giving Wrong Answers5 Dec 2022} Even OpenAI’s chief executive, Sam Altman, has conceded that “ChatGPT is incredibly limited, but good enough at some things to create a misleading impression of greatness.”^{2Vice (Motherboard)NYC Bans Students and Teachers from Using ChatGPT4 Jan 2023} A tool that sounds authoritative and is often wrong is hazardous to someone who cannot tell. It is also, for that reason, an unusually good teaching aid. Few exercises sharpen a pupil more than finding the errors in a fluent, confident essay. A school that blocks the bot gives up that lesson and keeps the temptation.

The third is that the ban sets a poor precedent for how institutions meet new technology. OpenAI says it does not “want ChatGPT to be used for misleading purposes in schools or anywhere else”, and that it is working on ways to identify AI-generated text.^{2Vice (Motherboard)NYC Bans Students and Teachers from Using ChatGPT4 Jan 2023} Detection may help, and teachers who suspect a pupil should certainly be free to ask for the work to be redone in class. But we suspect that schools which adapt their assessment will outlast those which merely block the device. Essays written at home were always the soft spot; the chatbot has only made that visible. Set more work on paper, hold more oral examinations, and grade the process rather than the product.

The strongest counter-argument is that a school system cannot be expected to hold a seminar on epistemology before it acts. A temporary block buys time to write guidance, and the department allows exceptions for schools that want to study the tool.^{1Chalkbeat New YorkNYC schools ban ChatGPT, writing tool that uses artificial intelligence3 Jan 2023} That is fair. Caution is cheap, and a pause of a term or two would be defensible. The trouble is that the stated grounds are permanent ones: the tool, Ms Lyle says, does not build critical thinking, full stop.^{1Chalkbeat New YorkNYC schools ban ChatGPT, writing tool that uses artificial intelligence3 Jan 2023} Nothing in that reasoning expires, so nothing in it suggests the block will be lifted when the guidance is ready. Temporary measures in large bureaucracies have a way of outliving the people who ordered them.

There is also a market lesson. Pupils will meet this software at university and at work, where employers will pay for people who can direct it and check it. Schools that teach those habits early will send out better graduates than those that teach a polite fiction. Banning the chatbot from the network teaches children only that the rules apply where the Wi-Fi does.