Portfolio item number 2
Short description of portfolio item number 2 
Short description of portfolio item number 2 
Published in Proceedings of the 33rd AAAI Conference on Artificial Intelligence, 2019
Classical NLP methods show how speech and voting records of U.S. senators differ in systematic ways.
Last updated 2022.
Exploring high-variance “trial balloon” projects can provide informational value even when they are ex-ante disliked.
Published in Economic Theory, 2023
In Bayesian persuasion, a Sender with maxmin preferences linearizes the prior.
Last updated 2025. Revision coming soon.
Partisanship causes more dissents on U.S. appellate courts, but those dissents do not always follow partisan lines.
Submitted, 2026.
We use large language models to extend standard political science codebooks to the universe of published U.S. Court of Appeals opinions (440,000 published opinions, 1892–2025).
Submitted, 2026.
Trump-appointed appellate judges do not produce more conservative case outcomes than other Republican appointees, but they do cast more conservative votes from the bench.
Graduate Student Teaching Assistant, MIT Department of Economics, 2021–2025