Scholars Debate Gatekeeping as arXiv Flooded With Papers Authors May Not Understand

Around September 14, 2024, a wave of AI-involved submissions to arXiv and other venues triggered intense academic debate. Machine-learning scholar Thomas Dietterich posted a long thread warning of a new trend: authors submitting papers they likely do not understand themselves, undermining arXiv's premise that human authors are accountable for content. He floated remedies including rejecting submissions from authors without publication records in the field, oral-exam certification, and a new mechanism called aiXiv. The discussion spread quickly, pitting governance advocates against anti-gatekeeping critics; arXiv itself has not commented.

Confirmed

• Dietterich's careful framing won support: the problem is not AI writing per se, but its combination with authors who do not understand and cannot take responsibility for the content (@brhkim).
• A proposal emerged for AI-detection firm Pangram to offer arXiv a free API that auto-desk-rejects submissions above an AI-generated threshold (@analisereal, relayed by @brhkim).
• @analisereal added that auto-rejection alone cannot fix the problem but removes incentives for cheap paper-milling, and suggested temporary bans for repeat offenders.
• @akpiper proposed an alternative: a "professional certification" mark distinguishing reviewed, understood submissions.
• Stanford professor Anshul Kundaje shared the thread and mocked rumors of arXiv introducing oral-exam-style review (relayed by @RishiBommasani).
• @RexDouglass relayed opposing views: if a paper is meaningful, valid, reproducible and readable, it should be judged on its merits—no one has the right to decide who "understands" a field.
• ObhishekSaha strongly objected that rejecting papers for insufficient author credentials is the worst form of gatekeeping, violating the universalism norm of Merton's ethos; AmirRPeimani also voiced strong opposition.
• Philosopher Seth Lazar announced the new journal Oxford Studies in Philosophy of AI and Computing, which will use conference sessions to screen out "meat robot" submitters.
• Nina Begus (@begusgasper) reflected that we are the last generation to write the scientific canon in a human way, and posterity will view us as ancient Egyptians viewed the pyramids.

Unconfirmed

• The oral-exam idea remains Dietterich's personal proposal; it is unknown whether arXiv will adopt it, and the Pangram auto-rejection and ban schemes remain community-level discussion with no official arXiv commitment.

Why it matters

• arXiv's moderation model rests on human author accountability; AI ghostwriting shakes that foundation, and unchecked low-cost flooding could dilute the credibility of the preprint ecosystem. The debate marks a shift from whether AI-assisted writing is allowed to how to verify authors understand and stand behind their work—oral exams, detection APIs, bans and certification marks may become routine tools, while critics like ObhishekSaha warn such mechanisms must not devolve into arbitrary gatekeeping that violates universalism.

2026-09-14 ~ 2026-09-15 · 11 related posts

• [source] AI-written papers flood arXiv: Dietterich proposes oral exams and an aiXiv venue for AI research — tdietterich · 2026-09-14
• [source] New Oxford journal adds conference stage to weed out authors who can't explain their AI-written papers — sethlazar · 2026-09-14
• Researchers propose auto desk-rejecting AI-generated arXiv submissions — analisereal · 2026-09-14
• [source] Follow-up: auto desk-reject plus temporary bans to curb AI-generated paper spam — analisereal · 2026-09-14
• arXiv sparks backlash over panels testing whether authors understand their own work — RexDouglass · 2026-09-14
• AI-written papers authors don't understand are flooding arXiv, warns Dietterich — begusgasper · 2026-09-14
• arXiv Sees New Trend of Submissions Whose Authors Likely Don't Understand Their Own Papers — _akpiper · 2026-09-14
• Should arXiv screen authors by credentials? Researchers clash over gatekeeping — anshulkundaje · 2026-09-14
• Researchers clash over arXiv 'oral exams' proposal amid flood of AI-written submissions — RishiBommasani · 2026-09-15
• arXiv moderation policy sparks backlash amid surge of papers authors may not understand — anshulkundaje · 2026-09-15

1 near-duplicate retellings: brhkim