Margaret J. Foster

Strategic Intelligence | Data Science | Emerging Risk

margaretjfoster.net | github.com/margaretfoster | [email protected]

Profile

Data scientist and former intelligence analyst developing methods to detect, measure, and interpret weak signals in opaque and strategically manipulated environments. My work combines machine learning, network science, simulation, text analysis, and qualitative domain expertise to identify latent change, emerging vulnerabilities, and structural risk. Current research spans national-security threats, illicit financial networks, critical-material supply chains, and organizational adaptation.

Core Capabilities: Theory-guided measurement from sparse information; interpretable machine learning; network science; latent-variable modeling; Bayesian inference; simulation and Monte Carlo validation; dimensionality reduction; text-as-data; LLM-assisted information extraction.

Professional Experience

University of Virginia

2024–Present

Data Science Researcher, National Security Data and Policy Institute

Remote · Charlottesville, VA

• Lead and contribute to interdisciplinary research developing quantitative tools for national-security problems characterized by incomplete, noisy, or strategically obscured data.
• Develop machine-learning, network-science, and simulation-based approaches for detecting structural vulnerability, hidden actors, organizational change, and potential substitution in complex systems.
• Lead research on critical-material and advanced-microelectronics supply-chain resilience, combining international trade data, network embeddings, observed disruptions, and qualitative validation to identify vulnerabilities and plausible alternative suppliers.
• Develop frameworks for determining when multiple imperfect signals can reliably recover otherwise unobservable states and processes, including simulation of shared and independent confounding, data sparsity, and incomplete observation.
• Contribute to interdisciplinary research on illicit financial actors, cryptocurrency transaction dynamics, state-backed networks, and AI-assisted extraction from large collections of public records.
• Translate technical findings into decision-relevant assessments, visualizations, briefings, and policy-facing research for national-security audiences.
• Collaborate across the National Security Data and Policy Institute and UVA's Biocomplexity Institute with political scientists, computer scientists, network scientists, and domain specialists.

SITE Intelligence Group

2007–2022

Analyst (2007–2008), Senior Analyst (2009–2014), Consultant (2014–2022)

Remote · Washington, DC

• Conducted open-source intelligence analysis of militant organizations focusing on leadership, strategic communications, and organizational behavior in rapidly evolving threat environments.
• Synthesized fragmented primary-source and open-source information into time-sensitive assessments of actors, incentives, organizational change, and emerging threats.
• Developed substantive expertise in interpreting strategic behavior when evidence is incomplete, adversarial, and intentionally manipulated.

Duke University

2023–2024

Postdoctoral Researcher

Durham, NC

University of North Carolina at Chapel Hill

2020–2023

Postdoctoral Fellow

Chapel Hill, NC

Selected Publications

Foster, Margaret J. “Growth Trap.” Studies in Conflict & Terrorism. 2026. Developed a theory of organizational change in armed groups, showing how shifts in rank-and-file preferences can reshape leadership behavior and organizational strategy. Adopted a theory-structured triangulation-based measurement approach, using multiple imperfect indicators to recover otherwise difficult-to-observe internal change. DOI: 10.1080/1057610X.2026.2707953
Role: Solo author

Foster, Margaret J. “Subject to Change.” Political Science Research and Methods, 2025. Addresses a persistent disciplinary measurement dilemma of how to conceptualize and capture dynamism and evolution in armed conflict actors. I introduce a measurement strategy, evaluate the face validity, and demonstrate scalability to a corpus of 258 militant groups with more than ten violent events from 1989–2020. DOI: 10.1017/psrm.2024.26
Role: Solo author

Alabsi Aljundi, Amro, Abhijin Adiga, Christopher Barrett, Margaret J. Foster, Brian D. Klahn, Dustin Machi, Achla Marathe, Philip B.K. Potter, Aaron Schroeder, Samarth Swarup, Anil Vullikanti, Mandy L. Wilson, and Madhav Marathe. “ Network-Based Active Learning for Identifying Illicit Actors in Financial Transaction Networks. ” Proceedings of the 25th International Conference on Autonomous Agents and Multiagent Systems (AAMAS), 2026. DOI: 10.65109/gock1684
Role: Conceptualization, framing and application to labeling clandestine data points.

Morucci, Marco, Margaret J. Foster, So Jin Lee, Kaitlyn Webster, and David Siegel. “ Measurement That Matches Theory: Theory-Driven Identification in Item Response Theory Models. ” American Political Science Review, 2024. Winner, 2025 Margaret Levi Award for the Advancement of Comparative Methodology. Introduced a novel Bayesian constrained item response theory model to measure latent concepts in observational data, validated through simulations, existing benchmarks, and applied to survey data. DOI: 10.1017/S000305542400039X
Role: Conceptualization, algorithm development, applications.

Johnson, Tana, and Margaret J. Foster. “ Rhetorical frames: a strategy of governmental power in international negotiations. ” Third World Quarterly, 2024. Uses text analysis to quantify conventionally weak states exerting rhetorical influence in a deliberative diplomatic setting. DOI: 10.1080/01436597.2024.2385103
Role: Conceptualization, measurement, data generation and cleaning.

Foster, Margaret J. “Gig Economy Insurgency.” Studies in Conflict & Terrorism, 2024. Draws on primary and secondary open-source evidence that militant groups have also adopted a “gig economy” model, proposes a typology of informal delegation, and develops predictions for consequences for conflict dynamics. DOI: 10.1080/1057610X.2024.2367256

Minhas, Shahryar, Cassy Dorff, Margaret J. Foster, et al. “ Taking Dyads Seriously: A Latent Network Approach for Global Politics. ” Political Science Research and Methods, 2021. Shows how a regression-based approach, the Additive and Multiplicative Effects (AME) model, can be used to account for inherent dependencies in dyadic observational data, correcting foundational specification errors in models of interrelations between actors. DOI: 10.1017/psrm.2021.56
Role: Implementation

Foster, Margaret J., and David A. Siegel. “ Pink Slips from the Underground: Changes in Terror Leadership. ” International Studies Quarterly, 2019. Develops a formal theory model of constituent dependencies that addresses an observational puzzle of why terrorist groups publicly fire wayward commanders, rather than kill them. DOI: 10.1093/isq/sqz017
Role: Conceptualization, model design, qualitative case development.

Selected Current Research

Foster, Margaret J. “Information Reconstruction.” In development; working paper available upon request. Develops a simulation-based framework for determining when multiple noisy and potentially confounded traces can recover an otherwise unobservable state, with particular attention to shared versus independent sources of error.
Role: Primary contributor — conceptualization, development, execution, drafting.

Foster, Margaret J., Nathan Timbs, Damian Boldt, Jonathan Hathaway, Phillip B.K. Potter. “Fabrication Geopolitics.” Working paper, available upon request. Examines persistent critical-material processing bottlenecks, geopolitical vulnerability, and the technical, economic, and political constraints on supply-chain diversification.
Role: Primary contributor — conceptualization, development, execution, drafting.

Foster, Margaret J., Nathan Timbs, Damian Boldt, Jonathan Hathaway, Phillip B.K. Potter. “ Never Waste a Good Crisis: Using Geopolitical Shocks to Validate Substitutability in Complex Systems. ” Working paper, available upon request. Develops and validates network-based measures for identifying replacement suppliers after observed disruptions in advanced-microelectronics precursor supply chains.
Role: Primary contributor — conceptualization, development, execution, drafting.

Foster, Margaret J. “ Scale and Stability: A Probabilistic Framework for LLM-Assisted Data Extraction. ” Develops a framework for ontology-constrained LLM extraction that treats parsing probabilistically and propagates extraction uncertainty through the resulting dataset. We apply the pipeline to cross-jurisdictional administrative records on public investment in critical materials across the EU, the US, Japan, and South Korea between 2011 and 2025.
Role: Primary contributor — conceptualization, development, execution, drafting.

Foster, Margaret J. “ Measurement for Opaque Systems: Multi-source Triangulation with Interpretable Machine Learning. ” arXiv:2602.00022 . Proposes a measurement framework for difficult-to-access contexts that uses indirect data, interpretable machine-learning models, and theory-guided triangulation to fill inaccessible measurement spaces.
Role: Solo author

Foster, Margaret J., David A. Siegel. “ IRT-M For Continuous, Ordinal, Categorical, and Mixed Data. ” Develops, implements, and validates code that extends the IRT-M supervised Bayesian IRT model beyond binary data, including extensive validation simulations.
Role: Conceptualization, development, simulations.

Johnson, Tana, and Margaret J. Foster. “ Beyond the 'BASICs': Upper-Middle-Income States and Obstructive Leadership in International Institutions. ” Working paper. Uses text-as-data and network analysis to study strategic contestation in multilateral negotiations; introduces a conceptualization of “obstructive leadership” to explain behavioral patterns of middle-power countries.
Role: Conceptualization, dataset development, measurement.

Schroeder, Aaron, Mandy L. Wilson, Galen S. Harrison, Dustin Machi, Brian Klahn, Samarth Swarup, Anil Vullikanti, Achla Marathe, Margaret J. Foster, Philip B.K. Potter, and Madhav V. Marathe. “ STDN-GEN: Rapid Synthesis of Layered Critical-Material Dependency Networks for Supply-Chain Sustainability Analysis. ” Working paper.
Role: Conceptualization, stabilization, validation.

Fleishman, Madeline, Damian Boldt, Nathan Timbs, Margaret J. Foster, and Philip B.K. Potter. “ Targeting State-Backed Networks: Evidence from Iranian Nuclear Science. ” Working paper. Uses network analysis and causal inference designs to estimate the impact of assassination programs on academic Iranian nuclear science output.
Role: Conceptualization and framing.

Huang, Fenix, Henning S. Mortveit, Margaret J. Foster, Philip B.K. Potter, and Christian M. Reidys. “ Dynamic Regimes in Cryptocurrency Transaction Data. ” Working paper. Develops and applies a novel information-theoretic model to identify data regimes, improving risk classification in a benchmark cryptocurrency dataset.
Role: Framing and application for observational data.

Selected Presentations, Panels, and Workshops

2026. Invited Panelist, “Critical Materials and Defense.” Governor's Annual Elements of Change: Lithium & Critical Minerals Summit, Reno, Nevada. Conference postponed due to the Hawk Fire.

2026. Panel Moderator, National Security Data and Policy Institute Conference, Resilient Supply Chains and the Defense Industrial Base.

2025. “Estimating Substitutability in Complex Systems Using Shocks.” Sunbelt Conference / International Network for Social Network Analysis.

2023. “Peace Operations Data, Analytics, and Action.” Invited workshop, United Nations Department of Peace Operations, New York.

2022. AI, NLP, and Text-as-Data Workshop. Prepared for United Nations Department of Peace Operations staff and partners.

2022. Working Session on Monitoring and Measurement of Peace. Principles for Peace, Geneva, Switzerland.

2022. “Domino Effect: Network Structure and Recruit Cohesion.” CASBS Seminar Series for the Organizations and Their Effectiveness Institute.

2019. “Frontiers in Organizational Studies.” Center for Advanced Study in the Behavioral Sciences, Stanford University.

Software and Research Infrastructure

IRT-M. Co-developer of an R package for theoretically identified, semi-supervised Bayesian dimensionality reduction. Available on CRAN: cran.r-project.org/web/packages/IRTM .

Additional research infrastructure includes simulation environments for multi-source triangulation, synthetic-data validation, network degradation and falsification tests, trade-network substitutability analysis, and LLM-assisted structured information extraction.

Education

2020

Ph.D., Political Science, Duke University

Political Methodology; Security, Peace, and Conflict

2011

M.A., Political Science, McGill University

2007

B.A. with Distinction in International Relations, University of Pennsylvania

Cum laude

Selected Awards and Recognition

• 2025 Margaret Levi Award for the Advancement of Comparative Methodology , for “Measurement That Matches Theory.”
• Carnegie Junior Scholar, Carnegie International Policy Scholars Consortium, 2016–2018.
• Honorable Mention, Best Poster Award, Political Networks Conference, 2017.
• Honorable Mention, National Science Foundation Graduate Research Fellowship Program, 2015.

Professional Service

Reviewer for American Political Science Review, Nature Human Behaviour, Conflict Management and Peace Science, and Review of International Organizations.

Communications Director, Political Networks APSA Section, 2020–2023; Program Committee member, Political Networks and Politics and Computational Social Science conferences.

Languages

French (Advanced)

Margaret J. Foster · Strategic Intelligence · Data Science · Emerging Risk