Background and Motivation
Interdisciplinary research is critical for solutions to complex societal challenges, and institutions have invested in a variety of programmatic models, from top-down research centers to faculty-driven collectives. Team science research has extensively documented the value of interdisciplinary collaboration, yet studies focus largely on formally structured teams. Far less is known about how large, self-organizing faculty networks function as engines of team science — particularly for early-career scholars navigating institutional barriers like siloed departments and publication-centric promotion criteria. Evaluating informal academic networks is itself a methodological challenge, as their outputs resist traditional metrics like publications and grant awards.

Research Questions
This study addresses two related questions: (1) How does a self-organizing interdisciplinary network develop connectivity and disciplinary diversity over time? and (2) What individual and structural factors enable deeper collaborative engagement within the network?

Case Study and Methods
We examine The STEAM Factory – a diverse and inclusive self-governing faculty network at The Ohio State University with over 200 members spanning 66 departments across 13 colleges. Founded in 2012 by a small group of early-career faculty and postdoctoral scholars, the STEAM Factory has grown into one of the most disciplinarily diverse academic networks at a major U.S. research university, operating with a flat organizational structure and without top-down programmatic directives.

We use a mixed-methods approach combining longitudinal network analysis with qualitative interviews. Self-reported collaboration data — spanning research, teaching, and outreach activities — were collected via Qualtrics surveys administered in 2018 (N=121; response rate=71%) and 2020 (N=160; response rate=72%). In ongoing work, we are adding data from the 2022 member survey to the analysis. Network visualizations and community detection algorithms (including Newman-Girvan edge betweenness, Greedy Modularity Optimization, Louvain, and InfoMap methods) were applied using the igraph package in R to characterize network structure and track changes over time. To explore factors associated with collaboration depth, we conducted eight semi-structured interviews with members stratified by their number of network connections and analyzed transcripts using fuzzy-set qualitative comparative analysis (FsQCA).

Key Findings
The network deepened substantially over the two-year period: total reported collaborations increased by 42% (from 136 in 2017 to 194 in 2019). Community modularity scores declined across all algorithms (from approximately 0.74 to 0.68), indicating that disciplinary communities became less siloed and more integrated over time, with 23 members serving as cross-community bridges by 2019. Members tended to be most actively collaborative within the first two to three years of joining the network. Qualitative analysis revealed that highly connected members were intrinsically motivated to seek out collaborations and engaged with both academic colleagues and public audiences. The network's neutral, self-governed space emerged as a critical structural feature enabling collaboration across disciplinary and rank-based hierarchies.

Figure 1: STEAM Network 2017 and 2019

Implications for Team Science
Grassroots models can cultivate robust interdisciplinary teams organically — without the constraints of project-specific mandates or senior leadership directives. The STEAM Factory offers a model for institutions seeking sustainable, inclusive team science ecosystems, and this study offers a methodological framework for evaluating informal networks where traditional output metrics fall short.