The unprecedented availability of high-quality data has revolutionized industries from healthcare to transportation, e-commerce, and beyond. In this talk, I will discuss a series of studies based on longitudinal data from a decade of interdisciplinary scientific conferences. In our first study, we developed a nonlinear dynamical model inspired by the physics of catalysis to predict how scientists form teams based on their interaction at conferences and showed that this conceptually simple model performs surprisingly well at predicting which participants self-assemble into teams. In follow-up work, we showed that team formation driven by formal interaction is similar at in-person and virtual conferences, but virtual conferences connect communities about half as much as in-person conferences. We found evidence that these effects persist beyond the end of the conference and may have effects on teams years later. Then, we extend the notion of time-varying interaction in groups beyond pairs using hypergraph methods and assess the predictive value of adding information on higher-order interactions for team formation. The talk will conclude with a discussion of ongoing work studying the dynamics of communities in the ARCH, a new, digital platform for scientific collaboration. These activities are guided by the overarching question: how can we design scientific ecosystems to optimally promote the innovations needed to meet global challenges?

Big Science initiatives like the ATLAS experiment at CERN exemplify the scale and complexity of modern collaborative research. With thousands of scientists and institutions from over 40 countries, ATLAS represents a global effort to uncover fundamental aspects of particle physics. In this work, we investigate the evolving collaboration patterns within ATLAS by constructing and analysing bipartite networks of authors and countries linked to their publications. Through this dual perspective, we uncover structural features such as modularity at the author level, and a clear nested pattern at the country level, each reflecting distinct organizational dynamics: modularity highlights the formation of cohesive working groups, driven by bottom-up interactions and stabilized by institutional continuity. Nestedness, on the other hand, underscores the stratified contributions of nations based on resources and expertise, revealing both strengths and vulnerabilities in the collaboration. Using percolation analysis, we assess the robustness of these patterns to perturbations, finding that modularity ensures resilience to individual turnover, while nestedness reveals fragility to the loss of key contributors. These findings shed light on the interplay between structural organization and dynamical stability in large-scale collaborations, offering insights for managing and optimizing similar scientific endeavours.

Most research focuses on the dynamics of social phenomena. Taking an interest in sociology raises the opposite question, that of its stagnation. Modern sociology has been affected by this problem since its beginnings at the end of the 19th century, and sociologists regularly express it in the form of a ‘malaise’. This expression is generally accompanied by a more precise questioning about the status of the discipline, about its capacity to be a ‘science’, and thus to find the terms of a unity around what makes its explanation. However, adopting a sociological perspective on such a question leads us to formulate a paradoxical question: can sociology explain why sociologists do not agree on their explanations?

In order to provide an answer to this question, my work focuses on the two central dimensions of its activity. Its aim: to respond to ‘social demand’; and its means: to adopt both the norms and the achievements of science. My work shows that not only does such a choice deprive sociologists of their object, but that it justifies and normalise the state of uncertainty that characterises them. By re-examining the aims and means of sociological activity in the light of science, it is however possible to defend the idea that it is possible to explain sociological stagnation in terms of the sharing and justification of a shared definition of its activity, that is problematic in terms of the aim of science.

States regulate the institutions of science, govern the dynamics of professionalization in scientific disciplines, and set ethical, theoretical and empirical standards on scientists’ work. They also orient research priorities through funding or the emulation of research in evidence-based policy. Through their expert bureaucracies, states also directly influence scientific knowledge through production, curation, interpretation and dissemination of research. Extant research thus documents various mechanisms through which states shape the content of science. Yet, researchers have seldom looked at the role of states, and especially expert bureaucracies, in shaping the structure and boundaries of scientific fields. Do states indeed make scientific fields? We explore this question through an empirical study of the U.S. Centers for Disease Control (CDC), an agency that has outsized prominence in the biological sciences both within the United States and globally. Our project analyzes the impact that CDC-published scientific articles have on the structure and substance of the subfields to which they contribute. We derive a corpus of 16,069 CDC publications from OpenAlex and identify the subset of all publications within the same scientific domain as each of them. Our analysis uses time-series cocitation networks within each of these domains to identify the effects of CDC publications on subfield structure. Preliminary findings suggest that publications from the CDC incentivize scientists to synchronize the framing of subsequent publications to those of the CDC’s articles, decrease the diversity of ideas in a scientific domain, and promote publications that are in line with the state-sponsored presentation.

Gender differences in scientific recognition remain an important topic. While authorship is a key form of credit, men and women may receive recognition differently. This study examines authorship and acknowledgment patterns across various disciplines to explore these differences. Our analysis shows that women are more often acknowledged than listed as co-authors, particularly in investigation and analysis roles. However, in collaborations between highly and less-cited researchers, highly-cited women tend to receive authorship more often than their male counterparts. These findings suggest that scientific credit is shaped by multiple factors, including gender, power, and perceived success. Understanding these differences can help promote fairer recognition practices in academia.