How Science Peer Review Fails and What We Can Do About It

Modern science's heavy reliance on customized, private computer code has rendered traditional peer review obsolete, as reviewers can no longer verify the underlying processes or detect hidden errors and fraud.

This systemic lack of transparency, coupled with a "publish or perish" culture that disincentivizes replication and code-sharing, has led to a growing "replication crisis" where scientific integrity is frequently compromised for career advancement.

Peer review was built for papers whose reasoning fit on the page, and almost nothing in science fits on a page anymore: roughly three million articles arrive yearly, most resting on private code nobody else can inspect. When statisticians requested the code and data behind 204 computational papers in Science, barely a quarter of the acquired results could be reproduced one study in PNAS. Meanwhile 2023 closed with more than 10,000 retracted papers, a record Nature ties partly to sham peer review Nature. Publishers pocketing an estimated ten billion dollars a year feel no pressure to change, and scientists earn no career credit for releasing code. December's retraction tally arrives next, when the field learns whether any of this landed.