The Race — Building Democracy in the Age of AI. People race toward a shared finish line.
The Race
← Trends

Supporting case

Scientific teams show a narrower division of tasks after ChatGPT's arrival

What happened

A November 2 preprint by Stefano Bianchini and Georgiana-Ioana Coroama examines contribution statements from roughly 120,000 PLOS ONE papers published in 2018–2025. Its event-study models compare task participation before and after late 2022, controlling for field and team characteristics. It finds fewer co-authors credited within several activities and narrower individual task portfolios, particularly in newly formed teams. Established collaborators show substantially less change. For example, writing participation in new teams is about three percentage points below its 2022 level by 2025.

The authors do not observe individual researchers' AI adoption. Contribution statements are self-reported and record participation rather than hours, effort or authority. One journal also limits generalization. These are associations around a shared technological turning point, not proof that AI eliminated scientific jobs or caused each observed change. Methods, results and limitations ↗.

For The Race, the useful mechanism is organizational: new collaborators may reorganize tasks more readily than established teams. It challenges the assumption that giving everyone the same tool automatically changes how institutions work. It does not yet establish smaller viable teams, a transfer of decision rights or an overall concentration/dispersion outcome.

Dates and implementation stage

Action/stage 2026-11-02; publication 2026-11-02

Supporting sources

Related trends