Breaking the Silos: How Institutional Homophily Governs Interdisciplinary Team Assembly in Health Research

How do research silos form, and how can we break them? Using an agent-based model extending team assembly theory, we simulate how tuning institutional homophily bridges basic biomedical scientists and clinical epidemiologists to foster translational health research.
David Fajardo Ortiz Oct 03, 2026

The Problem: Institutional Silos in Health Research

Bridging the gap between basic laboratory discoveries and clinical outcomes remains a fundamental challenge in modern biomedical science. At national research centers—such as Mexico’s National Institute of Geriatrics (INGER), established to address population aging—researchers operate within traditional departmental structures. Scientists are typically divided into three units: basic research (subindividual and cellular level) alongside clinical epidemiology and demographic epidemiology (individual and population level).
This departmental design fosters distinct scientific cultures with contrasting jargons, tools, and quality standards. An empirical co-authorship network analysis of INGER scientists (178 publications across 21 researchers) reveals stark disparities in team assembly strategies:
  • Basic Biomedical Scientists: Form larger research teams (average size of 7.48 members) that rely heavily on external collaborators. Only 22% of their co-authors are internal INGER colleagues, and just 4% of basic research teams contain two or more INGER scientists. Consequently, basic researchers remain scattered on the institutional periphery.
  • Clinical and Demographic Epidemiologists: Form smaller teams (average size of 4.78 members) with strong internal cohesion. On average, 45% of their members are internal colleagues, and 43% of teams incorporate two or more internal scientists. They form a dense central giant component anchored by highly connected academic leadership.
Without structural intervention, basic research insights remain localized, failing to percolate into clinical and population health applications.

The Mechanism: Extending the Team Assembly Model

To model how institutional rules shape collaboration topology, researchers extended the classic NetLogo Team Assembly Model developed by Bakshy & Wilensky (2007) and grounded in Guimerà et al. (2005). While the original framework models team formation through newcomer entry and incumbent memory in homogeneous populations, the extended model introduces two coexisting "breeds" of scientists—basic biomedical researchers and clinical epidemiologists—interacting under four structural probabilities:
  1. p (Incumbent Probability): The likelihood that a selected team member is an existing researcher (incumbent) rather than a newcomer entering the network.
  2. q (Incumbent-Incumbent Repeat Link): The probability that an incumbent selects a previous co-author.
  3. r (Incumbent New Link): The probability that an incumbent selects an existing institutional researcher without prior co-authorship.
  4. s (Newcomer Repeat Link): The probability that a newcomer joins a previous collaborator of the team's initiating incumbent.
To govern cross-field integration, the model adds an institutional homophily parameter (hom). Functioning as a governance dial, hom dictates the probability that researchers choose co-authors from within their own discipline versus across departmental boundaries. High homophily (hom → 100%) forces strict intra-disciplinary selection, reinforcing silos. Lower homophily (hom → 0%) encourages cross-field collaboration regardless of departmental affiliation.

The 3-Act Network Transition: Video Walkthrough

The following 75-second video walkthrough demonstrates how adjusting institutional homophily fundamentally reorganizes the collaboration architecture of a health research institute:

  • 0:00 – Act 1: Baseline Institutional Divide (hom = 79%)

    Video Caption: "0:00 – Baseline (hom = 79%): High departmental isolation with basic research clusters disconnected from the clinical core."

    At 79% homophily, the simulation accurately reproduces the empirical network structure observed at INGER. Clinical researchers coalesce into a dominant "giant component," while basic researchers remain fragmented in peripheral, isolated clusters.

  • 0:25 – Act 2: Multidisciplinary Shift (hom = 46%)

    Video Caption: "0:25 – Multidisciplinary (hom = 46%): Emergence of cross-field bridging co-authorships connecting basic and clinical hubs."

    When homophily is reduced to 46%, cross-departmental collaboration links begin to emerge. The network reorganizes into two major disciplinary hubs joined by narrow bridging co-authorships—a classic multidisciplinary configuration.

0:50 – Act 3: Translational Integration (hom = 14%)

Video Caption: "0:50 – Interdisciplinary (hom = 14%): Full network integration and structural blending across scientific cultures."

Lowering homophily to 14% triggers a percolation transition. The boundary between departments dissolves, absorbing previously isolated researchers into a single, fully integrated interdisciplinary giant component.

🧪 Try the Live Simulation Yourself

You can test these parameter transitions directly in your browser without installing any software:

👉 Assembling interdisciplinary teams

Preset Experiments to Try:

  1. Set hom = 79% to observe the institutional baseline (departmental silos).

  2. Set hom = 46% to observe the emergence of multidisciplinary bridges.

  3. Set hom = 14% to watch full interdisciplinary percolation

Policy Implications for Science Management

Our simulation results demonstrate that institutional network topology is highly sensitive to collaboration preferences. Modulating a single governance dial—institutional homophily—can shift an organization from isolated research silos to a fully integrated translational network.

However, reducing homophily in real-world institutions requires active policy intervention. Interdisciplinary collaboration is challenging because basic and clinical researchers operate with different jargons, methodological standards, and team assembly habits.

To promote translational science, research leaders can implement targeted organizational strategies:

  1. Institutional Consensus Workshops: Organizing structured strategic planning sessions to define shared institutional research priorities across departments.

  2. Ontology Engineering & Shared Jargon: Utilizing structured semantic tools and ontology engineering to help researchers across different fields build a shared vocabulary and common knowledge base.

  3. Simulation-Informed Governance: Using agent-based models as policy sandboxes to evaluate proposed funding mechanisms and collaboration incentives before implementing them institutionally.

By actively managing team assembly practices and lowering interdisciplinary barriers, research institutes can foster more resilient, collaborative networks capable of addressing complex population health challenges.

References

  • The Primary Research Paper:

    García-Peña, C., Gutiérrez-Robledo, L. M., Cabrera-Becerril, A., & Fajardo-Ortiz, D. (2019). Team assembly mechanisms and the knowledge produced in the Mexico’s National Institute of Geriatrics: A network analysis and agent-based modeling approach. Scientifica, 2019, Article 9127657. https://doi.org/10.1155/2019/9127657

  • The NetLogo Simulation Model:

    Cabrera-Becerril, A., & Fajardo-Ortiz, D. (2019, update in 2026) Assembling interdisciplinary teams. https://www.modelingcommons.org/models/Nkg_cU98cxwmalzhB-8Lo

  • Based on Bakshy, E. and Wilensky, U. (2007). NetLogo Team Assembly model. http://ccl.northwestern.edu/netlogo/models/TeamAssembly. Center for Connected Learning and Computer-Based Modeling, Northwestern University, Evanston

  • The Theoretical Foundation:

    Guimerà, R., Uzzi, B., Spiro, J., & Amaral, L. A. N. (2005). Team assembly mechanisms determine collaboration network structure and team performance. Science, 308(5722), 697–702. https://doi.org/10.1126/science.1106340[cite: 9, 10]