Scientific data, as a crucial research output, has now widely gained recognition for its academic and economic value (Greenberg, Wu, et al., 2023; Pasquetto et al., 2017; Wilson et al., 2014). Under the impetus of open science, the forms of data publication are becoming increasingly diverse and the content is continually enriched (Landi et al., 2020; Wittenburg, 2021). As the infrastructure for data openness and sharing is progressively established and refined (Benhamed et al., 2023; Greenberg, McClellan, et al., 2023), the issue of how to incentivize researchers’ willingness to open their data has become a new challenge for scientific data sharing in an open environment (Faniel & Jacobsen, 2010; Tenopir et al., 2011; Treadway et al., 2016).

In recent years, it has been recognized that appropriate and meaningful incentives are essential to capitalize on the promise of data sharing (Lo & DeMets, 2016) and that crediting data generators is key in this effort (Kalager et al., 2016). As stated by the International Council for Science (ICSU), ‘Scientists should be recognized and given credit for the scientific contribution of the data sets that they produce as well as for the analysis of those data’ (ICSU). Consequently, Barbara E. et al. proposed the designation of ‘data authors’ as an incentive for data sharing, with explicit identification in publications (Bierer et al., 2017). Notably, more than a decade ago, Jillian C.’s exploratory study found that for many participants, the term ‘author’ in the context of data was not fitting, raising the question: according to scientific researchers, is data something that can be authored (Wallis & Borgman, 2011)? We believe that this shift in perception is related to the development of data publication and the rise of data papers over the past decade. In the dissemination of data papers, in order to be cited as a data author, a person must have made substantial contributions to the original acquisition, quality control, and curation of the data, be accountable for all aspects of the accuracy and integrity of the data provided, and ensure that the available data set follows FAIR Guiding Principles (Bierer et al., 2017). However, accurately defining data responsibilities to clarify the identity of data authors is a current challenge.

In data-intensive scientific research activities, establishing the identity of data authors requires that their contributions to the data are quantifiable and evaluable. With the proliferation of author contribution statements, various contributor role ontologies and taxonomies (CROTs) have emerged (Hosseini, Colomb, et al., 2023), offering standardized lists of roles or terms to designate individuals’ contributions to research. Among these, CRediT (NISO), as a standardized method for describing author contributions, has been vigorously promoted by numerous journal publishers and widely adopted since its introduction in 2014. It is also gradually being applied to data papers to describe the contributions of authors in the production and publication process of scientific data.

In this research, we employ CRediT to examine the division of scientific labor in a sample of data papers from Data in Brief, exploring how research contributions are allocated in data-intensive scientific activities. More specifically, we first explore the intercorrelations among the 14 CRediT roles to assess the utilization of CRediT in data papers. We also consider the relationship between the division of scientific labor and the size of the author team, as well as the discipline of the research. Finally, we investigate the correlation between key authors in data papers, such as corresponding authors and first authors, and the CRediT roles they undertake.

This study employs CRediT to explore the division of labor in data-intensive scientific research activities, represented by data papers. According to our result, the pattern of labor division in data articles is different from traditional research papers in many ways. For instance, in data papers, the core contribution of ‘conceptualization’ is most strongly correlated with the contribution of ‘methodology’, while it is ‘funding acquisition’ with research articles in the RCT field (Zhang et al., 2019). However, the roles that play a significant part in data paper collaborations show no significant difference from those in research articles: almost all articles involve roles such as ‘conceptualization’, ‘writing – original draft’, and ‘writing – review and editing’ (Lariviere et al., 2021). Furthermore, some roles exhibit high autocorrelation and lack distinct differentiation, suggesting that they could be considered for consolidation within the context of contributor roles in data papers.

Data-intensive science is a collaborative endeavour, and our research indicates that as the number of co-authors increases, there is a corresponding rise in the number of individuals taking on roles such as investigation, validation, and formal analysis. In contrast, the number of individuals assuming core roles like conceptualization, methodology, and Writing – original draft, or managerial roles such as project administration, funding acquisition, and resources, remains relatively stable. This suggests that the primary driver of team size expansion is the need for practical tasks such as data handling and investigation, while the number of scientific leaders remains scarce (Robinson-Garcia et al., 2020). As Larivière et al. have posited, ‘the bureaucratization of science can be considered as an inevitable consequence of the ubiquity of collaborative science’ (Larivière et al., 2015).

The publication of data papers reflects the work patterns and research cultures across various disciplines. Different academic fields exhibit varying degrees of concentration in the use of CRediT roles, with data papers in the social sciences generally showing higher concentrations of CRediT role usage compared to those in the natural sciences. We suggest that the differences in research methodologies, types of data, and the nature of research questions are the primary causes of this phenomenon. For instance, data papers in the social sciences more prominently represent the ‘paper’ dimension, with authors tending to focus on the ‘conceptualization’ role, while data papers in the natural sciences emphasize the ‘data’ dimension more, with authors concentrating more on ‘investigation’ and ‘methodology’.

The attribution of academic credit is one of the key concerns for researchers, with a common understanding that corresponding authors and first authors make significant contributions and play major leadership roles (Bhandari et al., 2014; Perneger et al., 2017; Teixeira da Silva, 2021; Yang et al., 2017). The statistical analysis indicates that in data papers, corresponding authors often undertake more leadership and coordination tasks and are more likely to be involved in key roles such as ‘conceptualization’, ‘funding acquisition’, and ‘supervision’. These roles pertain not only to the initial design and theoretical construction of the research but also encompass the supervision of the research process and financial support. The first authors, on the other hand, exhibit higher engagement in roles like ‘writing - original draft’ and ‘data curation’, which are directly related to the implementation of the research and the accuracy of the data. Comparatively, corresponding authors are more likely than first authors to take on any given role, implying that they act as versatile players within the team (Lu et al., 2020) and are more likely to be the corresponding authors of a data paper. Furthermore, our research also reveals that the more roles an author takes on, and their involvement in key data processing roles such as ‘data curation’, ‘software’, and ‘visualization’, the more it aids in the author’s ranking in terms of by-line order.

Our research indicates that some technical contributions related to data processing are clearly very important for the attribution of credit to authors of data papers. However, in the CRediT, the connotations of roles such as ‘data curation’ encompass both work requiring profound professional background knowledge, such as data annotation, and technical tasks like data cleaning and maintenance. The roles of technical contributions (Smith, 2023) and human intervention are not adequately differentiated and described. This could lead to a situation where some authors’ significant contributions are not recognized as they should be, while others enjoy excessive credit due to their role designation. This situation could be more pronounced in data papers, ultimately affecting the fairness of academic credit attribution for data papers.

CRediT, as a standardized method for describing author contributions, aids in clarifying the specific contributions of each author and determining the data responsibilities of different authors in data papers. However, the content standards and organizational forms of data papers differ from those of traditional research articles (Callaghan et al., 2012). Data papers focus more on the collection, processing, and presentation of data, some of which do not have direct corresponding categories within the 14 CRediT roles. The majority of data papers use only about 9 roles to describe the work conducted, indicating that CRediT has limitations in fully capturing the research workflow. Some roles, such as ‘validation’ and ‘software’, which ensure data quality and reuse value, are missing in most data papers. This leads to an underestimation or neglect of the actual significance and unique contributions of these roles in research work, affecting a comprehensive understanding of the entire research process. This uneven usage also implies that CRediT fails to accurately and consistently provide all important types of contributions when describing the research work in data papers (Alliez et al., 2020; Fitzgerald et al., 2020; Matarese & Shashok, 2019).

Over the last few decades, data-intensive scientific research has become increasingly prevalent, with some of its labor division characteristics reflected through the publication of data papers. Our study arrives at the following main conclusions:

Data papers provide an excellent window into studying data-intensive scientific research activities, and CRediT offers a useful framework for characterizing data-centric scientific workflows, but it requires refinement to reflect the characteristics of data papers and the diversity of research work more comprehensively. Developing a taxonomy of contributor roles specific to data papers—DP-CRediT (Data paper-CRediT)—could be a good option. For instance, it might be beneficial to add specific roles such as ‘data collection’ and ‘metadata management’ to accurately reflect the contributions of these key steps in data papers. Moreover, the importance of technical work such as software development and data processing is increasingly recognized. Alliez et al. argue that since software development in research involves ‘significant innovation’, there is a need for appropriate human intervention and qualitative information to ensure accurate reporting of contributions (Alliez et al., 2020). Building on the existing roles of ‘software’ and ‘formal analysis’, it could be considered to further refine technical contributions by introducing roles such as ‘algorithm development’ and ‘data engineering’ to describe technical work more comprehensively. Adding roles like ‘critical analysis’ or ‘data interpretation’ could emphasize the role of human intervention in data analysis and interpretation, ensuring a balance between technical contributions and human judgment. While refining roles, it is also necessary to maintain consistency among existing CRediT roles. Some CRediT roles describe project-level tasks while others describe paper-level tasks(Hosseini, Gordijn, et al., 2023), which could lead to an imbalance in the allocation of academic credit.