DM: We now have the perfect storm: Paper mills using genAI. The technology has become cheap and fast. We also have lazy reviewers that use AI to do the work. Gen AI enables not just the fabrication of fraudulent papers; it enables making up entire research pipelines. #WCRI2026
Post
Replies:
5
DM: We work on soiutions to detect synthetic Western blots.
We ran a Kaggle Competition for copy/move detection of real retracted data.
www.kaggle.com/competitions...
Recod.ai/LUC - Scientific Imag...
Next up: Stefan Stender with "AI Misuse to Scale Production of Human Health Research Manuscripts" Conventional epidemiology was in a crisis in the 1990s: associations appeared robust but were non-causal due of confounders. Mandelian randomizations was better. wcri2026.org/program/ #WCRI2026
SS: Now: Genome Wide Association Studies, there are 100,000 GWAS datasets as well as user-friendly R-packages. These packages produce publication-ready figures. This facilitates the mass production of "Mendelian randomization" studies, most from China, flooding the literature. #WCRI2026
SS: We see advertisements for increasing your CV by writing Mendelian Randomization papers based on a simple tool, and we tested that LLMs can easily produce MR papers as well. These formulaic types of papers are also highlighted in the @cosig.net@bsky.brid.gy guidelines: osf.io/2kdez/files/... #WCRI2026
SS: This is not limited to MR papers - it is starting to become a widespread problem with open-access datasets. We need to shift from article count to research quality. Papers can be produced in 20 min - they lose their meaning. We need to shift away from counting paper metrics.
