How will science cope with a flood of plausible hypotheses, generated papers, synthetic data, and results that few humans can independently verify?
Messeri & Crockett's warning: AI can produce more science while producing less understanding — epistemic monocultures where everyone's hypotheses come from the same models and nobody can check the volume.
View on the map → · Open in Browse →
The flood became measurable: a BERT screen flagged nearly 10% of 2.6 million cancer papers as sharing paper-mill patterns, Nature documented tens of thousands of 2025 papers with likely AI-fabricated references, and the first publisher-deployed review-fraud detector went live.
The model flagged 261 245 of 2 647 471 papers (9.87%)
Screen of 2.6M cancer papers flags 9.87% as sharing text patterns with known paper-mill work, including in high-impact journals.
Tens of thousands of 2025 papers may contain AI-fabricated references; surveys the editorial fixes being proposed.
The first publisher-deployed detector for duplicated or suspicious peer-review reports.