Collaborative Networks

Finding Colleagues and Emerging Subdisciplines via Systematic Network Analysis of Funding Awards

Scientific teams often expand by connecting researchers interested in similar topics from across institutions or disciplines. Expansion can be especially important when teams look to pursue funding by which to scale previous work into research centers or hubs. For researchers in nascent subdisciplines or in remote institutions, however, finding potential colleagues can be difficult and overly dependent on chance, stalling the growth of subdisciplines and privileging those at large urban universities. In this paper, we present protocols for systematically finding potential colleagues based on publicly available funding awards. As a proof of principle, we focus on the US National Science Foundation's award database, and in particular on it's Broadening Participation (BP) portfolio, a collection of awards each of which aims partly to increase the number of underrepresented minorities in STEM fields. We focus on the emerging sub-discipline of BP education programs for grades K-12. From the lens of the funding agency, we demonstrate the scope of that subdiscipline, its primary players, and we indicate the extent to which those awards reveal collaboration networks. Those in the subdiscipline can use the results to systematically find potential collaborators and tap into extant networks. This work is based on computer assisted, but relatively low-tech, techniques to create large databases, find a subset of awards relative to research questions, find research outputs, and visualize collaboration networks. The protocols used to conduct this work can be redeployed to address many questions about funding agencies portfolios.

Patterns of newly formed interdisciplinary collaborations over time during an immersive training experience

Background: Understanding how team and disciplinary communication patterns predict work quality could help interdisciplinary project teams to optimize their performance.  We continue to examine the evolution of project-related communication among newly formed interdisciplinary teams embedded for one week at the annual NIH mHealth Summer Training Institute. The aim is to characterize communication in new interdisciplinary teams and to explore how patterns might predict performance outcomes.

Objective: To characterize communication trajectories over time both within and between teams and within and between disciplines engaged in mobile health. Based on 2017 findings, we hypothesized that (1) greater within team communication early in the week would predict better oral presentations; (2) greater between team communication later in the week would predict better written capstone projects; and (3) data scientists would emerge as the most central cross-disciplinary connectors.

Methods: 2018 fellows (n=29) were split into Teams (n=5) and assigned mentoring faculty from different universities and disciplines (data science [computer science/engineering/informatics]; medicine/nursing;  behavioral science; public health). Teams were given an immersive training experience and charged with developing a capstone project to address a health challenge. Requirements were to produce a final oral presentation and written grant proposal evaluated by independent reviewers. At the end of days 1, 3, and 5 of the 5-day training, fellows (nodes) indicated with whom they discussed their project (edges) since the prior assessment. Team and disciplinary network density were calculated as the ratio of observed edges to potential edges within and outside the team and the discipline. Internal and external network densities were examined to characterize emergent communication patterns and to explore predictors of oral and written project evaluations.   

Results:  As the week progressed, and teams approached the project deadline, the density of project-related communication increased both within and between teams and within and between disciplines.  Whereas 2017 teams with the best rated oral and written presentations showed distinctive evolution of project-related communication, emergent communication patterns were more homogenous in 2018.  An exception was that the poorest performing team showed a mid-week decline in project-related communication both within and between teams.  Finally, examination of disciplinary interactions showed that data scientists again served as essential hub connectors in 2018, closely followed by medical professionals.

Implications:  In both 2017 and 2018, the density of project-related communications increased both within and across teams and disciplines as project deadlines neared.  Greater homogeneity in network communication trajectories among 2018 teams made it challenging to discriminate high- from low-performing teams, failing to reproduce 2017 predictors of project performance.  Data scientists emerged as key integrators of communication across interdisciplinary mHealth teams in both years.  Here, we demonstrate a feasible methodology for collecting and analyzing team communication networks. Prediction of interdisciplinary team performance would benefit from a larger sample of teams, data collection in other contexts, and exploration of other potential predictors.

Network Improvement Communities and Organizational Change: Considerations in Developing and Leading NICs

Scholars from multiple fields and knowledge domains have been called to work together by National Science Foundation (NSF) and other funding agencies to address some of the most complex challenges facing society. Network improvement communities (NIC) are recently identified as a means to champion these efforts. Funders of cross-unit collaborations are interested in knowing how to bring diverse faculty together to address challenging problems including how to form scholarly communities, break down structural barriers to working across boundaries, construct rewards and incentives to support these efforts, foster organizational change, and consider longer-term sustainability – which is especially pertinent if an initiative is funded. Outcomes generated through collaborative alliances such as network improvement communities (NICs) are prominent areas of research, but the process by which these groups are formed is largely understudied, especially at the postsecondary level. This presentation focuses on two research questions – (1) How do postsecondary NICs form, and (2) In what ways does the NIC assist members in facilitating change efforts on their campuses?

Data were gathered from an NSF funded NIC that is in its third year of existence. As the evaluation team of the NIC, data was collected through participant observation, survey, and individual participant interviews. As evaluators, we are specifically studying the evolution of the NIC as an organization, including ways in which trust, communication, shared norms and goals are developed, how dissonance and failures are accounted for, and how learning occurs at both the individual member and NIC level. As the NIC continues to evolve, we better understand its role in facilitating organizational change on member campuses even while it is not actually the agent of change. Although there appear to be some barriers at the institutional level (in the network under study) that remain unchanged by the NIC, the NIC does act as a centralized learning hub that advances and expedites the learning process of the group, which can result in institutional change. The information presented during this presentation identifies specific processes and elements to success and sustainability such as leadership, strategy, communication, engagement, institutional infrastructure, data collection and analyses, building capacity and the possible emergence of network capital that can assist in developing protocols to support institutional change for NICs. It also shows some of the pitfalls of human resource organizations, such as NICs, that must be addressed in order to be truly more than resource shells and “group work.” Consideration is also given to the NIC as a learning organization.

Beyond providing current knowledge from an evolving network improvement community, the presentation intends to encourage discussion about the utility of NICs as facilitators of change as well as to raise questions for future research that deepen our understanding of these emerging collaboratives.

Investigating Collaborative Processes of Research Teams through Social Network Analyses

In the evaluation of team science, questions may arise as to which team members or groups of members have access to and control information, as well as if collaborations among team members produce increased scholarly productivity. Network analyses within a social network analytic framework offers one method for mapping influence and scholarly productivity within a team project. We present the results of a brokerage analyses and exponential random graph modeling (ERGM) from an evaluation of an NIH-funded multidisciplinary, cross-institutional biomedical research center. The evaluation assessed whether groups within the center were functioning according to their primary intended role (i.e., administrative, research) to facilitate research and promote the sustainable development of the center, as well as if collaborations among team member produced increased evidence of scholarly productivity in the form of publications and presentation of scientific work. Data on collaborative processes were collected annually from questionnaires administered to all members of the center. Three years of evaluation data are presented. Cross sectional brokerage and longitudinal ERGM analyses provided evidence for team functioning.

Social network analyses were performed to address the following research questions:

1. What is the emergent community network structures of collaborative engagement among center members?
2. What brokerage processes between core center areas drives the emergence of the observed center community structure?
3. How do these brokerage processes align with the predefined roles of center cores areas?
4. Does collaboration among people from different core areas produce increased scholarly productivity among center members?

The evaluation team designed a survey to gather information about team work patterns. This information was used to identify (a) how information flows amongst members of a research center, (b) whether organizational structures within the center functioned to control information, and (c) if the roles of individuals and organizational structures shifted over time and lead to increased scholarly productivity.

First, modularity analyses were conducted on bipartiate network graphs to determine if community clusters self-organized around a priori organizational structures. Next, brokerage analyses were conducted at the group level. A brokerage score for a given node is the number of ordered pairs having the appropriate group membership brokerage relationship. Aggregate scores can be computed for defined groups within a network as well as at the network level. Expectations and variances of brokerage scores given the size and density of a network were also be computed. Analyses were conducted for 3 years of census networks and compared descriptively over time. These data were then analyzed using longitudinal ERGM’s to determine if increased ties between members with different core function in the collaborative lead to increased productivity on behalf of individuals. The results provide insight into the emergent structure of the research collaborative and the influence of structural relationships on scholarly productivity and knowledge production.

SciTS Presentation: Investigating Collaborative Process of Research Teams through Social Network Analysis

Understanding Scientific Collaboration through the Sequence of Authors in the Publication Bylines and the Diversity of Collaborators

In science of team science, it is critical to investigate the patterns of scientific collaboration and how these patterns result in different impacts. In this research, we investigate the relationship among (a) scientific collaboration, (b) the sequence of authors in the publication bylines, and (c) the diversity of their collaborators. The diversity of collaborators is quantified with two dimensions, namely topic and impact diversities. Using the ArnetMiner dataset containing ACM-indexed publications in computer science, we find that the following two patterns tend to lead higher-impact scientific publications: (1) greater topic diversity of collaborators plus more tendency to work as leading authors (including first and/or corresponding authors); and (2) less topic diversity of collaborators plus less tendency to work as leading authors. Meanwhile, from the perspective of impact diversity, the results of our empirical study show that authors who work as more leading authors and collaborate with less impact diversity researchers have tendencies to receive more citations than those with collaborators with greater impact diversity.

SciTS Presentation: Investigating scientific collaboration through the sequence of authors in the publication bylines and the diversity of collaborators