How can you adequately measure the individual contribution of a researcher if the scientific works that make up their curriculum are the result of collaborations and are also signed by other colleagues? In search of answers to this question, which appears frequently in assessment processes, researchers from the Institute of Mathematics and Computer Science of the University of São Paulo (ICMC-USP) and the University of Indiana, in Bloomington, USA, have examined 243 million scientific articles from 2.6 million coauthors stored in the Microsoft Academic Graph database. To adjust the target, they disregarded papers with more than 10 coauthors, in order to not contaminate the analysis with the work of large research consortia, and they only selected authors from among those who had published at least 10 articles and received at least 200 citations, excluding those that were at the very beginning of their career. For each author from the database, the most important collaborator was identified — and these collaborators were analyzed individually.
The results showed to what level research collaborations have peculiar dynamics in different fields of knowledge and revealed curious data about the influence of extremely productive academic leaders among colleagues with whom they publish scientific articles. One reference point was the figure of “prolific collaborators,” authors whose names frequently appear on the list of authors for articles and whose work receives a greater volume of citations. To assess the influence of these leaders, the study analyzed what happened with the productivity metrics of the other authors, measured by citations and indicators that join the quantity of articles and their repercussion, such as the h-index, in two different scenarios: with and without the articles made in cooperation with the prolific collaborators. “There were extreme cases in which an author from the medical field counted 4,000 citations in articles they coauthored, but when the papers produced with the prolific collaborators were excluded, only 25% of the original citations remained,” states Diego Raphael Amancio, of ICMC-USP and lead author of the study, published in the scientific journal Scientometrics. “This could be a sign of dependence, but also of the success of the collaboration. Other information will need to be analyzed to see whether the relationship is of dependence, such as what was the contribution of the authors, for example.”
In humanities subjects, it was observed that the influence of this elite is more restricted, especially since there is not a strong tradition of publishing in partnership. In engineering, computer science, physics, medicine, biology, and chemistry, the weight was significantly greater, with the prolific collaborators responding to around half of the publications and citations from their specialty. In chemistry, the index that measured the influence on the total number of citations reached 50% of the total. In materials science, it was 66%. For Amancio, the results may be useful for understanding how collaborations function and for identifying characteristics that define researchers capable of producing science autonomously.
In the assessment of computer scientist Jesús Mena Chalco, a researcher from the Federal University of ABC who did not participate in the study, the work led by Amancio explores two questions already highlighted in the literature but that have not yet been fully quantified: the relationship between the intensity of the collaboration and the researcher’s field of work and the size of the influence of someone more experienced in a research environment. “In my opinion, this last question, which covers the impact of a more collaborative researcher on the scientific production of their colleagues, is the most important,” he says. According to Chalco, this evidence could help in the discussion about scientific policies for engaging senior researchers in challenging projects. “Brazil could benefit from a routine and systematic policy of bringing together the ‘oldest’ with the ‘youngest.’”Republish