Research

My research focuses on developing quantitative methods for empirically hard questions in fields of political science in which data is scarce and causal analysis is hard. Examples include the study of civil unrest, institutional change, and democratization. I approach these problems by developing methods to easily collect and analyze new non-standard types of data, such as image, audio and video, and by developing methods to assess the robustness and sensitivity of causal data analyses in these contexts, where traditional identification is hard. In addition to this, the tools I develop are applicable widely in political science, and are designed with the goal of enabling a large and diverse group of researchers to perform high-quality data analysis, independently of their resources.

Peer-reviewed Publications

• Multi-Task Learning Improves Performance In Deep Argument Mining Models With Amirhossein Farzam, Shashank Shekhar, and Isaac Mehlhaff. Proceedings of the 11th Workshop on Argument Mining ( ACL Argmining 2024). (To appear) Arxiv

• Anti Political Class Bias in Corruption Sentencing With Luiz Vilaça and Victoria Paniagua. American Journal of Political Science (2024) SSRN Published Version Replication

• Measurement That Matches Theory: Theory-Driven Identificationin IRT Models. With Margaret Foster, Kaitlyn Webster, So Jin Lee, and David Siegel. American Political Science Review (2024) Published Version Arxiv Replication

• A robust approach to quantifying uncertainty in matching problems of causal inference. With Md. Noor-E-Alam and Cynthia Rudin. INFORMS Journal on Data Science (2022) Published version Arxiv Replication

• FLAME: A Fast Large-scale Almost Matching Exactly Approach to Causal Inference. With Tianyu Wang, M. Usaid Awan, Yameng Liu, Sudeepa Roy, Cynthia Rudin, and Alexander Volfovsky. Journal of Machine Learning Research (2022). Published version Arxiv Python Package R Package

• AME: Interpretable Almost Exact Matching for Causal Inference. With Haoning Jiang, Thomas Howell, Neha R. Gupta, Vittorio Orlandi, Harsh Parikh, Sudeepa Roy, Cynthia Rudin, Alexander Volfovsky. Neural Information Processinf Systems Competitions and Demonstrations Track (NeurIps Demo track 2021) Published version

• Adaptive Hyper-box Matching for Interpretable Individualized Treatment Effect Estimation. With Vittorio Orlandi, Sudeepa Roy, Cynthia Rudin, and Alexander Volfovsky. Uncertainty in Artificial Intelligence (UAI 2020) Published version Arxiv Replication R Package

• Almost-Matching-Exactly for Treatment Effect Estimation under Network Interference. With M. Usaid Awan, Vittorio Orlandi, Sudeepa Roy, Cynthia Rudin, and Alexander Volfovsky. Artificial Intelligence and Statistics (AISTATS 2020) Published version Arxiv Replication

• Interpretable Almost-Matching-Exactly With Instrumental Variables. With M. Usaid Awan, Yameng Liu, Sudeepa Roy, Cynthia Rudin, and Alexander Volfovsky. Uncertainty in Artificial Intelligence (UAI 2019) Published Version Arxiv Replication

• Shuffling the House? The Profile of Candidates and MPs in the 2015 British General Election. With Chrisa Lamprinakou, Rosie Campbell, and Jennifer vanHeerde-Hudson. Parliamentary Affairs (2016) Published version

Work Under Review

• Matching Bounds: How Choice of Matching Algorithm Affects Treatment Effect Estimates. With Cynthia Rudin (R & R at Journal of Politics) Arxiv

Working Papers and Drafts

• Model Complexity for Supervised Learning: Why Simple Models Almost Always Work Best, And Why It Matters for Applied Research. With Arthur Spirling. Draft

• A Double Machine Learning Approach to Combining Experimental and Observational Data With Vittorio Orlandi, Harsh Parikh, Sudeepa Roy, Cynthia Rudin, and Alexander Volfovsky. Arxiv

• Matched Machine Learning: A Generalized Framework for Treatment Effect Inference With Learned Metrics With Cynthia Rudin and Alexander Volfovsky. Arxiv

• dame-flame: A Python Library Providing Fast Interpretable Matching for Causal Inference. With Neha R. Gupta, Vittorio Orlandi, Chia-Rui Chang, Tianyu Wang, Pritam Dey, Thomas J. Howell, Xian Sun, Angikar Ghosal, Sudeepa Roy, Cynthia Rudin and Alexander Volfovsky. Arxiv Python Package

• A Fast, Cost-Efficient Image Annotation Algorithm withan Application to Analyzing the Visual Components of Protests. Draft

• Credible Assumption Mixtures: Combining Point and Partial Identification Assumptions for Interpretable Bayesian Causal Inference. Draft

• Political and Economic Determinants of Illicit Financial Flows. With Pablo Beramendi, David Dow, and Erik Wibbels (Draft available upon request)