120k-paper study: LLMs made research teams more specialized, with new teams changing most
JMateosGarcia · x · 2026-11-05
A preprint by Bianchini & Coroama (University of Strasbourg) uses CRediT statements from 120,000 PLOS ONE articles (2018-2025) to study how ChatGPT changed scientific division of labor. Three findings: (1) participation in research tasks fell below prior trends, especially conceptualization, execution, and writing; (2) researchers became more specialized, with narrower task portfolios and less overlap with co-authors; (3) changes concentrated in newly formed teams, while established teams showed organizational inertia. The authors also propose a CRediT-AIR scheme annotating AI involvement (0-5 scale) per research activity, and speculate specialization may raise efficiency but reduce learning, particularly for early-career researchers.
Related event: Study of 120,000 papers finds LLMs make scientific teams more specialized(6 posts)→
More from AGI Musings
- Artists on craft in the AI age: marble-carving robots and film-generating models — every · 2026-11-05
- Training AI beyond language models: a practical session on RL environments — every · 2026-11-05
- Why won't mainstream media take the AI and digital culture beat seriously? — kyliebytes · 2026-11-05
- OpenAI's Greg Brockman: AI will make real strides curing diseases in 12-18 months — inductionheads · 2026-11-05
- Podcast: AI-to-medical acceleration Hot or Not — rare disease to revenue cycle — zakkohane · 2026-11-05
- Design essay argues the profession's crisis isn't AI, but its loudest players — round · 2026-11-05