Peer Review Without Peers

What's happening

The International Conference on Machine Learning rejected 497 papers, roughly two percent of submissions, after catching their authors using AI to generate peer reviews of other researchers' work. Conference organizers embedded hidden watermarks in papers distributed for review. When a reviewer fed a paper to a language model, the watermark triggered telltale phrases that exposed the fraud. ICML operates on a reciprocal review policy: every submitting author must also review others' papers. More than half of researchers in a 2025 survey admitted to using AI in peer review despite policies prohibiting it. "The thing we must protect most actively is our trust in each other," the organizers wrote. Marie Souliere of Frontiers publishing called it evidence of "a research community in need of clear guidance on responsible AI use." The conference will be held in Seoul in July.

What the text says

Proverbs 11:1A false balance is an abomination to Yahweh, but accurate weights are his delight.

The Hebrew word mirmah translated as "false" carries the meaning of deceit, treachery, fraud. A "false balance" is not merely inaccurate. It is deliberately rigged to give the appearance of honest measurement while producing a predetermined result. The word for "abomination" (to'evah) is the strongest term of moral disgust in the Hebrew Bible, used elsewhere for idolatry and injustice that corrupts the social fabric.

The marketplace of ancient Israel depended on scales. Grain, silver, and oil were measured by weight. The entire system of economic exchange required that the instrument of measurement could be trusted. When the scales were rigged, the transaction could look exactly like an honest one. Buyer and seller would go through every normal motion. The fraud was invisible at the point of exchange. It only became visible in its effects: the buyer received less than what was weighed, and the community's confidence in honest dealing eroded from the inside.

Proverbs treats the integrity of measurement as a theological issue because measurement is how a community establishes shared truth. When the instruments lie, everything built on them is compromised.

The reflection

Five hundred researchers built systems capable of generating language indistinguishable from human thought, then used those systems to avoid the work of actually thinking about someone else's research. The scales were false. The watermark caught what human readers could not. Proverbs locates the danger precisely: the corruption of measurement corrupts the community that depends on it. Peer review exists because science requires trust between strangers. When the instrument of trust is itself automated, the question is no longer about one conference. It is about whether any institution built on good faith can survive the tools designed to simulate it.