Not yet human-reviewed
This source record was written automatically by the research pipeline and has not yet been reviewed by the maintainer or integrated into a topic page. Treat it as a provisional summary and verify claims against the original source.
Scientific exploration, collaboration and labor division in the large language model era
Citation. Zheng, X., Hong, X., Liu, J., & Ni, C. (2026). Scientific exploration, collaboration and labor division in the large language model era. arXiv. https://doi.org/10.48550/arXiv.2607.20923
Summary. This preprint links PubMed Central and OpenAlex publication histories for 775,323 scientists with CRediT contribution data from 137,120 multi-author papers to describe post-2022 changes in scientific exploration, collaboration, and team role differentiation. It reports temporal associations with the LLM era rather than causal effects of LLM adoption.
Key claims
- From 2022 to 2025, the average number of primary fields in scientists’ publications rose from 2.3 to 2.8, and the authors estimate that level as 15.6% above continuation of the pre-2023 trend.
- In 2025, 35.5% of authors published in at least one previously unentered field, compared with 30.4% in 2022; the authors estimate this as 23.2% above the pre-2023 trend continuation.
- Scientists’ collaborator networks became more interdisciplinary after 2022, while research interdisciplinarity was less closely tied to collaborator disciplinary diversity among authors with stronger AI-writing signals.
- CRediT reports after 2022 showed narrower contributor role sets, less role overlap among coauthors, increased software and validation roles, and decreased conceptual and managing roles.
Evidence & limitations
- The study combines PMC full text and OpenAlex histories with a CRediT-role sample concentrated in PLOS and selected PMC-indexed journals, so its empirical coverage is primarily biomedical and biomedical-adjacent research.
- The AI-writing measure is a population-level textual proxy, and the descriptive design cannot observe private LLM use or identify LLM adoption as the cause of observed changes; post-2022 patterns may also reflect pandemic-era collaboration, funding, and broader changes in science.