Modeling scientific workforce dynamics using social network analysis

The scientific workforce requires teams to solve the most critical intellectual and social problems that confront us today. Scientists and inventors are embedded in self-organizing communities, where they share ideas and act both as critics and fans for each other. Recent research has shown that team collaborations, a growing trend across all disciplines, yield publications with higher intellectual impact than single researchers; and, the careers of young scientists are influenced by relationships with others in the community. Furthermore, we have found differences in the networks of women and minorities that explain some of the disparities that exist in these subgroups. Thus, we propose to develop a systems-based approach to studying scientific workforce dynamics that models the mechanisms of how new collaborations form and how these influence both the effectiveness of teams and the career trajectories of individual scientists. Obtaining the data needed to test these models may seem to be a formidable challenge. However, through prior projects, we have already brought together a unique collection of longitudinal datasets, linked at the individual person level, which will be utilized for this new study: On a national scale, PubMed (publications), NIH ExPORTER (grants), USPTO (patents - US Patent and Trademark Office), NPPES (health care providers - National Plan & Provider Enumeration System), and BoardEx (company directors and executives) provide data about individuals and teams both in academia and in industry. On a local scale, within Harvard University, we have collected detailed career data on 25,000 faculty across multiple disciplines, including sensitive information (e.g., race/ethnicity, time to promotion, grant application review scores, etc.) that are typically much more difficult to obtain. The national and local data are complementary, enabling models at different scales. This project will be undertaken by computer scientists and a behavioral and social scientist at Harvard University, who recently completed an NIH-funded project to study workforce inclusion and diversity, and social scientists from the Science of Networks in Communities (SONIC) lab at Northwestern University’s School of Communication, who are leaders in the use of Social Network Analysis (SNA) to model the socio-technical motivations of collaboration. Three specific aims are planned: (1) Develop empirically validated theoretical models that predict how teams form within the scientific workforce. We will use Exponential Random Graph Modeling (ERGM) to test Multi-Theoretical Multi-Level (MTML) models for the emergence of networks. (2) Determine how the assembly mechanisms of scientific teams influence their diversity and efficacy. We will model the composition of teams (e.g., gender, race, education, etc.) as well as a variety of productivity measures (e.g., citation counts and ability to obtain funding). (3) Determine the influence of a scientist’s collaborators on his or her career trajectory. We will incorporate SNA centrality metrics into models that predict advancement and retention in the scientific workforce, stratified by gender and race.