Mathematicians can't boycott their way out of AI, but they're right to try

Is it a guild reflex when a profession says no to the machine that wants its job? Sometimes. But when the profession is mathematics, the one discipline where "trust me" is not an acceptable argument, we should at least hear the objection out.

Here is what happened. As t3n reports, more than 1,000 mathematicians have pledged to give up AI entirely through the newly founded Association for Human Mathematics (AHM). Its stated mission is to protect mathematics as a human endeavor and to keep the community independent from corporate interest. The sharpest criticism is aimed at companies like OpenAI. According to The Decoder, the AHM accuses OpenAI of disregarding core norms of scientific research by publishing AI-generated proofs, and it urges mathematicians to stop working with the company. Euronews reports that OpenAI answered the backlash with a donation of five million.

The objection is about norms, not nostalgia

The easy reading is that mathematicians are afraid of being automated. We think that misses what the AHM is actually saying. The complaint, as the group frames it, is that a company published a large batch of proofs the community never asked for, and the community is now expected to digest them. The AHM's own line, as Euronews summarizes it, is that mathematicians did not commission this work.

That matters because mathematics runs on a social process. A proof counts when other people read it, check it and put their names behind it. In our view, a lab that publishes results at industrial volume risks shifting the cost of checking onto the community's referees. That is a fair grievance, and a donation doesn't answer it. Writing a check to the community you just burdened looks more like public relations than like a peer-review system.

A pledge of abstention is a weak lever

Now the uncomfortable part. A pledge is a statement of values, and a good one. But it is a poor tool for steering a technology that other people will keep building and using.

There is even a definitional problem. One outlet notes that the debate reaches as far as the question of what counts as AI use in the first place. Is a chatbot that explains a lemma off limits? A search engine with machine-learning ranking? A pledge that can't draw that line cleanly will probably be hard to enforce and easy to mock. That is a prediction on our part, not a documented outcome.

Then there is the classroom, where abstention already looks like a losing game. As t3n reports, Richard Borcherds of UC Berkeley no longer sees the point of grading his students' homework, because AI use has become unavoidable. A Hochschule Darmstadt survey found that 92 percent of respondents in Germany use tools like ChatGPT at least occasionally. That survey isn't about mathematics specifically, but we suspect the next generation of mathematicians is learning with these tools too, whatever the senior researchers pledge. A profession that opts out of the conversation will not stop that. It will just have less say in how it unfolds.

The real fight is over proof, credit and who checks

If we were advising the AHM, we would push it to use its 1,000-plus signatories as leverage for rules rather than abstinence. Three questions are worth demanding answers to:

• Verification: what is the standard for accepting a machine-generated proof, and who bears the cost of checking it?
• Credit: how are human and machine contributions disclosed, so that authorship still means something?
• Independence: how does the community avoid depending on one company's goodwill, or its donations?

None of these can be settled by staying out of the room. They are settled by journals, departments and funding bodies writing norms before the flood arrives. Ironically, the AHM's strongest card is the very thing it says it defends: mathematicians are the people best qualified to say what counts as a valid argument.

Our verdict

The pledge is not a guild reflex, or at least not only one. It is a rational defense of a fragile social process against what the AHM sees as a company that ignored the community's norms. As a signal, it works, and it should make labs think harder about how they release results.

As a strategy, it falls short. Mathematics will not be saved by individual abstinence, and a purity test gets harder to keep every semester. Our recommendation to mathematicians, and to the journals that serve them: keep the pledge as a moral anchor, but spend the real effort writing the rules for verification and credit. If you sign and then leave those rules to others, the humans in "human mathematics" will have lost the argument they cared most about.