Exploring the social organization of the sciences

Since Price presented his postulates on the shift to Big Science (1963), many have attempted at describing the social dynamics surrounding the development of scientific breakthroughs and the implementation of policies directed at fostering a productive research environment. Decades of work on interdisciplinary knowledge creation has advanced our understanding on knowledge-related dynamics across disciplines (Gibbons et al., 1994). Their challenges have been well documented in scientific collaboration studies (Bolduc, Knox & Ristroph, 2023; Klein, 2014). These address the dynamics of the inter-discipline, pointing to differences in language, research and team norms, and other disciplinary characteristics for working across boundaries (Leahey, Beckman & Stanko, 2017). These disciplinary distinctions are acknowledged, but not systematically examined.

From a theoretical perspective, Richard Whitley (2000) built a framework describing social and intellectual organization across fields based on two key concepts: mutual dependence and task uncertainty. While the former refers to the need for external validation, the latter focuses on labor distribution. Whitley suggests that different fields will have varying degrees of dependence and task uncertainty, proposing seven typologies of working structure. The dynamics of the research process are affected by myriad factors. Whether the research is more computational in nature, lab or field-research based, and reliant on specialized equipment or facilities are factors that drive the organization of scientific activities. Where a scientific team is based and where the scientists work introduce institutional, regulatory, and other factors that shape the nature of work.

Empirical studies on the social composition and dynamics of research activities have primarily focused on issues such as team size, power dynamics, collaboration or team composition and structure. This is due to the difficulty to examine the internal dynamics of team science at a large scale. These studies tend to rely on bibliographic data, applying bibliometric indicators based on author position, and combining it with individual traits such as gender or past trajectory. Science teams challenge current notions of credit in the scientific ecosystem (Walsh & Lee, 2015), as it becomes increasingly difficult to distribute it based solely on authorship position (Biagioli & Galison, 2013). In the last decade, the expansion of contribution statements across publication records allows for the first time to explore the relation between author order, time size and distribution of tasks, based on the self-reported statements of authors themselves. Benefiting from these novel data, we ask: how does intellectual leadership and research task distribution vary across disciplinary areas?

Data and Methods We analyze a set of almost 700,000 publications using the Contributor Role Taxonomy (CRediT) based on a set of Elsevier and PLOS journals to study the organization of the sciences across fields (González-Salmón et al., 2024). CRediT is increasingly being used by journals to catalog author contributions. Co-authors indicate their specific contribution out of 14 possible roles.

Building upon Whitley’s framework, we map the levels of task uncertainty and mutual dependence across and within fields. Considering that a paper reflects the outcome of a collective endeavor, we distinguish between scientific collaborations and highly stratified and organized team science with explicit structures. We apply agglomerative hierarchical clustering to publications by field and characterize the resulting clusters based on the team size and two measures of IDR, based on references and based on authors (Pinheiro et al., 2021).

Preliminary and Expected Findings So far we have analyzed the field of Medicine, but will extend our analysis to all fields prior to the conference. After applying bootstrapping and using different resolutions, our initial results suggest that there are four collaboration structures. These show different organizational structures related to differing modes of research. However, in terms of IDR, we find no meaningful differences across groups.

Cluster 1. All team members contribute to most tasks, with a moderate contribution on administration (e.g., funding). Papers are authored by a low number of researchers (avg. team size: 4.7). Cluster 2. Teams are characterized by a high distribution on technical tasks with a shared contribution on conceptual tasks. (avg. team size: 5.9). Cluster 3. Teams are larger with a higher distribution of labor for all tasks. (avg. team size: 7.5) Cluster 4. High distribution of tasks, although most members participate on the Review & Edit contribution (avg. team size: 8.2).

We will expand our analysis to all fields of research, exploring contribution co-occurrence by fields, and adding further variables such as seniority, gender, collaboration type and funding (based on funding acknowledgements).

Next steps The results of our research contribute to the understanding of collaboration and team dynamics in publishing activities. Because CRedit asks co-authors to indicate their role in the initiation, conduct, and communication of the research, it enables us to examine these activities across disciplines. These data greatly extend what has been possible to surmise from author order, but have limitations (Sauermann & Haeussler, 2017). There may be different approaches to how author contributions are reportedb However, given the size of our dataset, we can discern organizational structures across and within disciplines to understand the internal dynamics of team science.