In 2000, 1% of cancer papers were likely fraudulent. In recent years: more than 15%.
A new study flagged 250,000 cancer research papers as likely produced by "paper mills" — companies that mass-produce fake studies and sell authorships. The detection tool? An AI trained to spot patterns in titles and abstracts.
Here's the problem: even at 91% accuracy, we can't manually verify a quarter-million papers. We don't have enough integrity specialists. We can automate detection, but we can't automate trust.
And then there's the 4% false-positive rate. That means thousands of legitimate researchers potentially flagged as fraudsters. Careers damaged. Reputations questioned. All because an algorithm — not yet peer-reviewed itself — made a probabilistic guess.
The Deeper Problem
Step back further. Paper mills exist because someone is buying. Researchers facing "publish or perish" pressure. Institutions counting papers instead of reading them. A system that rewards volume over rigour.
We're building better fraud detectors. But we're not asking why the fraud market is booming in the first place.
What gets measured gets gamed. And right now, we're measuring the wrong things.
Low-quality papers are flooding the cancer literature — can this AI tool help catch them? (Nature) →
Originally published on LinkedIn · View discussion →