Return of the Solo Author: The Changing Division of Labor in Science in the Age of Generative AI

Jul 16, 2026

Modern science shifted from individual work to team production, improving productivity through the division of cognitive labor across specialized contributors. Generative AI may seem to extend this pattern by lowering research costs and enabling larger collaboration, while also substituting coauthor tasks and weakening collaboration needs. Analyzing over 300 million works across 26 fields, the study finds solo authorship decline halted and partially reversed after ChatGPT’s late-2022 release, varying by field and narrowing solo papers toward computational topics.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages37

Reading Time1 h 33 min

Abstract

Introduction

Two competing views on generative AI and labor division

Results

Solo authorship declines halt or reverse with ChatGPT

Method overview

Data source, coverage variants, and pre/post split

arXiv :2607 . 10780v1 [ cs .CY] 12 Jul 2026 Return of the solo author: The changing division of labor in science in the age of generative AI Akira Matsui Center for Computational Social Science, Kobe University, Kobe, Hyogo, Japan [amatsui@rieb.kobe-u.ac.jp](amatsui@rieb.kobe-u.ac.jp) July 14, 2026 Abstract Modern science has experienced a long shift from individual work to team production. Generative artificial intelligence (AI) might appear to extend this trajectory by lowering research costs and enabling larger-scale collaboration. Yet if tasks once performed by coauthors can be delegated to AI, the same technology may also weaken the need for collaboration in parts of the research process. Here, we examine this tension by moving beyond average team size and focusing on the solo-authored tail of the author-count distribution. Analyzing over 300 million works across 26 fields, we find that the decades-long decline in solo authorship halted and partially reversed with ChatGPT’s public release in late 2022 . We also reveal that this is an uneven phenomenon: it is strongest in fields where coauthors’ work is more readily replaceable, and weak or absent in fields that depend on physical collaboration. At the individual level, the recovery is not explained by the entry of new researchers or by changes in field composition. Instead, the break appears among authors who had written only with others, including those with no prior solo publications, and among longestablished authors as well as newcomers. Their solo papers stay close to their own coauthored work while narrowing in scope and shifting toward computational topics. Because a solo paper is work without credited human coauthors, this study offers an empirical probe of how generative AI can substitute for scientific labor, and evidence of a reconfiguration of cognitive labor within papers rather than of team size. Keywords: Science of science, Scientific collaboration, Authorship, Large language models, Artificial intelligence The division of labor is one of the most basic mechanisms for raising productivity. As Adam Smith described with the pin factory, dividing a single task into several specialized steps, each carried out by a different person, dramatically increases total output. Science is no exception. What science has divided, however, is not the labor of producing things but cognitive labor [1] . By forming research teams, researchers can divide their cognitive workload [2, 3], a necessity in scientific research that has grown ever more complex and specialized. As instruments, data, and collaborations have grown in scale and complexity, the cognitive labor behind a single paper has been divided among ever more specialized contributors [4, 5] . The well-known rise of team science is reflected in the decades-long increase in the average number of authors per paper [6, 7, 8, 9], and such teams can result in impactful outcomes [2, 10] . If team science reflects a deepening division of cognitive labor, how does generative AI affect that division? There are two views, and they make opposite predictions. The first is the acceleration view. LLMs have been argued to lower the cost of producing papers, increasing the number of coauthors and encouraging gift authorship [11] . On this view, generative AI extends the division of labor: more contributors join each paper, and the upper tail of the author-count distribution grows. The second is the substitution view, which focuses on the tasks allocated to individual contributors. Research tasks such as writing, data preparation, and statistical analysis have conventionally been divided among human coauthors [12] . LLMs now perform many of these tasks, from writing [13] to coding and data analysis [14, 15] . On this view, tasks in the division of labor shift from humans to LLMs, and researchers can complete work that once required collaborators on their own. This would imply an increase in solo authorship, that is

FAQs

What question does the study address about generative AI and scientific work?

It tests whether generative AI changes the division of cognitive labor in science by examining the solo-authored tail of the author-count distribution rather than relying only on average team size.

What empirical pattern is found after ChatGPT’s public release?

The decades-long decline in solo-authored papers halts and partially reverses beginning with ChatGPT’s late-2022 release, with the effect persisting in subsequent monthly analyses.

Why is the change in solo authorship uneven across fields?

The reversal is strongest in fields where coauthors’ work is more readily replaceable by AI, and weak or absent in fields that rely on physical collaboration.

Jul 16, 2026

Upload by River Wang