AI will reorganize science. Will research remain a human enterprise?

Summary

PNAS Perspective giving the systemic argument a name: a coevolutionary loop in which manuscripts are written to satisfy AI evaluators, those manuscripts become training data, and human input progressively thins - 'epistemic autophagy'. Identifies proxy optimisation over truth-seeking, reward hacking (citing the July 2025 hidden-prompt arXiv case), accumulating verification debt as production outruns human evaluative capacity, and model collapse. Concludes science risks becoming internally consistent yet cut off from the world it explains. The best single conceptual framing for an IS paper on AI-driven research pipelines.

Contribution

Queued — not yet read.

Method

Queued — not yet read.

Relevance to RISE

Surfaced by the 2026-08-29 literature sweep as a high-priority candidate for the RISE knowledge base (angle: autonomous AI scientists and their evaluation). Verified against: PMC full text (PMC13438523) opened 2026-08-29 (2 authors, PNAS 123(31), 2026-07-29, full argument read); PNAS site itself returns HTTP 403

Critique / open questions

Queued — not yet assessed.

Key quotes

—