A new platform gathers mathematicians’ views on AI through an evolving, anonymous survey open to eligible researchers.

Why ask mathematicians now?
AI-assisted solutions to famous mathematical problems have sharpened questions about how the technology may affect research and teaching. Futures of Maths, a new community platform, is asking research mathematicians to weigh in and help shape the questions under discussion. That gives the project a broader purpose than taking a snapshot of opinion: it’s also testing which questions the community thinks deserve attention.
The distinction matters. A survey can measure responses to a fixed list, but that list already reflects choices about what counts as important. By letting participants propose statements, Futures of Maths gives researchers a way to influence the agenda as well as respond to it. Whether that produces a more useful picture will depend on who takes part and which topics they choose to raise.
A survey that grows with its participants
Participants rate short statements about maths and AI, ranging from strong disagreement to strong agreement, or skip them. They can flag statements they consider especially important and see how other respondents have rated each one. The survey can be completed over multiple visits, which may make it easier to take part without setting aside a long block of time.
That format gives respondents several ways to engage: they can register a view, identify an issue as significant, or leave a statement unanswered. Those choices shouldn’t be treated as interchangeable. A skipped question may reflect uncertainty, lack of relevance, or simply a decision not to respond; the platform’s description doesn’t say how it’ll distinguish among those possibilities in its findings.
Researchers can also submit statements of their own. A moderator reviews each submission before it goes live for ratings, allowing the collection of questions to develop with community input. Moderation should help keep the survey usable, though the criteria for approval will shape which ideas make it into the shared discussion.
That’s a consequential design choice. An open prompt can surface concerns a survey’s organizers didn’t anticipate, while review can limit duplication or material that doesn’t fit the project. The balance will be visible in the final set of statements: which questions were proposed, which were approved, and which perspectives those questions invite.
Who can take part
Participation is open to people aged 18 or older who’ve published at least one article in a mathematics journal within the past five years. The journal must be indexed by zbMATH Open, and recently accepted articles also qualify. These rules define a specific group of respondents; they don’t, by themselves, establish that the group represents mathematicians more broadly.
The simplest verification route is signing in with ORCID. The qualifying article must appear on the person’s public ORCID record and have been added by Crossref. Articles entered manually don’t count. If Crossref has requested permission to add a work, the researcher can approve that request in their ORCID inbox and make the record public.
Researchers whose work is accepted but not yet published, or who can’t verify through ORCID, can request a case-by-case review using an institutional email address. That route acknowledges that publication records don’t always fit a single automated check. It also puts some weight on the review process, so clear and consistent handling will matter to people whose records need a closer look.
The organizers describe the criteria as a practical proxy for active research mathematicians, chosen for simplicity, objectivity, and ease of verification. They’re also inviting suggestions for alternatives that could reach a broader, more representative group while retaining those qualities. This is more significant than it looks: participation rules don’t just screen respondents; they influence which experiences and professional stages appear in the results.
If you’re working in this space, the eligibility rule is part of the research question, too. A recent journal publication is easy to explain and check, but it may leave out people whose mathematical work or teaching gives them a stake in AI without fitting that record. (And this is the part most people overlook.) The organizers’ request for alternatives signals that they recognize the tradeoff, even as they start with a workable boundary.
Privacy, return visits, and published findings
Eligibility checks take place on a separate website. The survey doesn’t receive participants’ names, email addresses, or ORCID identifiers. Each person gets a private return link at registration; it can’t be recovered or resent, so participants should save it.
Personal questions under “My session” are optional, and responses can be withdrawn at any time. Together, those features give participants control over whether to provide additional information and whether to keep their answers in the project. The separate eligibility check is also a meaningful part of the design: it aims to verify who qualifies without passing those identifying details to the survey itself.
There’s a practical cost to the private return link. If a participant loses it, the organizers say it can’t be sent again, which could make a multi-visit survey harder to complete. Anyone planning to return will need to keep the link somewhere accessible and private.
The team plans to publish findings once it’s collected enough responses, with an update on the Proofs and Prompts blog. The organizers are asking researchers to share the platform widely, particularly with colleagues whose perspectives are underrepresented in conversations about maths and AI. They haven’t specified a response threshold in the information provided, so readers will need to look to the promised update for details on when the results are considered ready to report.
What the project could show
Futures of Maths is trying to capture professional judgments at a moment when AI-assisted mathematical work is prompting questions about research and teaching. Its value will depend not only on the ratings, but also on the statements people submit, the mix of respondents, and how the organizers explain the limits of the sample. A set of opinions can make disagreements visible; it can’t, on its own, settle what AI will change or how quickly.
And yet, a community shaped survey can still be useful if it makes its boundaries clear. Readers will want to know which statements drew strong agreement or disagreement, which were marked important, and whose views are represented in those patterns. The platform’s invitation to share the survey with underrepresented colleagues points to that last question as an active concern, rather than a detail to address after publication.
For now, the project’s most concrete contribution may be the process it’s establishing: researchers can rate questions, propose new ones, and later examine the collected findings. A starting point, not a verdict.
Discussion
Sign in to join the discussion.