A growing body of evidence shows that collaborative teams and communities tend to produce the highest-impact scientific work. This paper proposes a new method to (1) Identify collaborative communities in longitudinal scientific networks, and (2) Evaluate the impact of specific research institutes, services or policies on the interdisciplinary collaboration between these communities. First, we apply community-detection algorithms to cross-sectional scientific collaboration networks and analyze different types of co-membership in the resulting subgroups over time. This analysis summarizes large amounts of longitudinal network data to extract sets of research communities whose members have consistently collaborated or shared collaborators over time. Second, we construct networks of cross-community interactions and estimate Exponential Random Graph Models to predict the formation of interdisciplinary collaborations between different communities. The method is applied to longitudinal data on publication and grant collaborations at the University of Florida. Results show that similar institutional affiliation, spatial proximity, transitivity effects, and use of the same research services predict higher degree of interdisciplinary collaboration between research communities. Our application also illustrates how the identification of research communities in longitudinal data and the analysis of cross-community network formation can be used to measure the growth of interdisciplinary team science at a research university, and to evaluate its association with research policies, services or institutes.

No abstract

OBJECTIVES/SPECIFIC AIMS: We aim to leverage our analysis of the scientific collaboration network at a research university to design an innovative pilot program and foster scientific productivity. We test the impact of creating a new collaboration in a research community, which decreases the average network distance and accelerates the diffusion of information and expertise among the community’s investigators. METHODS/STUDY POPULATION: We mapped the whole network of co-authorship on publications and co-participation on extramurally awarded grants at the University of Florida (UF) between 2013 and 2015. We used network science methods to identify research communities of investigators who have consistently worked together and/or have other collaborators in common with at least one researcher based in the UF Health Science Center. We selected pairs of communities with (i) similar productivity levels, research interests, and network structures and (ii) no research projects in common. Communities in each pair were randomly assigned to a treatment or control group. In each treatment community, we selected 1 pair of investigators who had not collaborated in the past 3 years and whose connection would maximally reduce average network distance in the community. The pair was provided with an economic incentive to collaborate for the submission of a CTSA pilot proposal. RESULTS/ANTICIPATED RESULTS: We successfully identified 15 pairs of treatment/control communities. In each of 8 treatment communities, a pair of potential collaborators agreed to participate in the intervention. DISCUSSION/SIGNIFICANCE OF IMPACT: Network-informed Clinical Translational Science Awards (CTSA) pilot programs can identify research communities and create innovative collaborations. Statistical experiments can establish the programs’ causal effects on scientific productivity.

Rapid innovation in science and technology has led to the development of new fields that transcend traditional disciplinary boundaries. Previous studies have retroactively examined the emergence of these fields. This paper outlines a mixed method approach for using network ethnography to identify emerging fields as they develop, track their evolution over time, and increase collaboration on these topics. This approach allowed us to simultaneously analyze organizational trends and gain an understanding of why these patterns occurred. Collecting ethnographic data throughout the course of the study enabled us to iteratively improve the fit of our models. It also helped us design an experimental method for creating new teams in these fields