Research

My research develops new methods for causal inference in panel and time series settings, combining Bayesian machine learning, micro-level data, and large language models to answer questions that conventional approaches cannot. I apply these methods to understanding how international organizations build state capacity in fragile states, where administrative and remote-sensing data are often sparse, noisy, or incomplete. Rather than treating imperfect data as a limitation, I treat it as the methodological problem worth solving, drawing on computational social science, fieldwork, and formal theory to recover credible causal evidence from difficult observational settings. Together, my work develops tools for the challenging data environments that characterize much of the social sciences.

This research has appeared in AISTATS, where my coauthors and I introduced a Gaussian process framework for estimating time-varying treatment effects in panel data. We are extending this work in a methodological paper introducing Gaussian process models to a broader social science audience.

Publications

A Multi-Task Gaussian Process Model for Inferring Time-Varying Treatment Effects in Panel Data (with Yehu Chen, Jacob Montgomery, and Roman Garnett) Proceedings of the 26th International Conference on Artificial Intelligence and Statistics (AISTATS), 2023.

Working Papers

Building State Capacity Locally: International Interventions, Delegation, and Local Governance in Fragile States (Job Market Paper) Selected presentations: the 2026 Society for Political Methodology Annual Meeting; the 2024 American Political Science Association Annual Meeting; and the 2024 International Studies Association Annual Convention.

A Gaussian Process Framework for Structured, Flexible, and Interpretable Machine-Learning Models (with Yehu Chen, Jacob Montgomery, and Roman Garnett) Presented at the 2024 Society for Political Methodology Annual Meeting; and the 2023 American Political Science Association Annual Meeting.

Why not both? Combining human and LLM labels in event data Presented at the 2025 American Political Science Association Annual Meeting; and the 2025 Society for Political Methodology Annual Meeting.

Beyond Peacekeeping: The United Nations Development Programme, Peacebuilding, and Violence Mitigation (Awarded Best Poster from APSA at the 2023 conference; Revise and Resubmit) Selected presentations: the 2023 American Political Science Association Annual Meeting; and the 2023 Society for Political Methodology Annual Meeting.

Border Fortification and Trust in Political Institutions (with David Carter, Beth Simmons, and Michael Kenwick) (Under review) Selected presentations: the 2026 European Political Science Society Conference; and the 2024 International Studies Association Annual Convention.

Works in Progress

• Hierarchical Partial Pooling and Difference-in-Differences (for the 2026 American Political Science Association Annual Meeting)
• United Nations Development Programme Complete Project Dataset
• Lights Out: The USAID Shutdown as a Natural Experiment in Nighttime Light Validity