Research

My research sits at the intersection of social movement studies, political sociology, and computational social science. I am broadly interested in how collective action unfolds over time, and particularly in how movements cohere, how they fracture, and how the cultural materials that sustain them shift in response to political and social change. Across my projects, I use computational methods and large-scale data to ask questions about meaning, contention, and power.

The LGBTQ Movement After Victory

My dissertation asks what happens to social movements after they win. Taking the American LGBTQ movement after the 2015 Obergefell v. Hodges decision as a case study, I examine how victory reshapes the cultural fabric of a movement: its coalitions, its framing and messaging, its collective identity, and the stories it tells itself about its own past.

The project draws on two original datasets. A dataset of nearly 600,000 Facebook posts from LGBTQ organizations spanning fourteen years allows me to trace how movement framing and coalitions shift over time. A parallel dataset of interviews with LGBTQ activists across the country and observations in activists’ meeting spaces allows me to investigate how meaning is made and remade within the movement. Across these datasets, I use a mix of computational, quantitative, and traditional qualitative methods to study how the movement’s cultural infrastructure evolved through and after the fight for marriage equality.

A core argument of the dissertation is that victory constitutes a singular crisis point for social movements. In its wake, movements must renegotiate collective identities, frames, coalitions, and political objectives — often simultaneously and under conditions of considerable internal disagreement and outside pressure form opponents. The wake of victory is a period of uncertainty about what the movement is, what it should priorities, and who it speaks for. The dissertation pays particular attention to how activists shape and deploy collective memory in these renegotiations, making sense of the recent movement past in order to stake claims about its present and future.

Settler Colonialism and Canadian Law

Together with Kimberly Huyser and Mary Jessome, this project explores how settler colonial governance has been reflected in Canadian legal, legislative, and regulatory text. We are building a computational pipeline for analyzing Canadian legal and legislative documents from the nineteenth century to the present, using natural language inference to examine how Indigenous peoples, Black Canadians, and issues regarding them appear (or fail to) in the legal record over time.

Previous work in this area has largely relied on close readings of limited subsets of legal text. This project asks a complementary question: using computational tools that can operate at scale, what patterns in the legal treatment of Indigenous and Black Canadians become visible across the full breadth of the documentary record that smaller analyses might not capture?

This project is in its early stages. Initial results are expected in late 2026.

What Do Language Models “Know” About the Past?

Together with Laura Nelson, this Schmidt Sciences-supported project investigates whether large language models can serve as instruments for historical social research. Historical perspectives are, by definition, inaccessible to modern-day researchers. We cannot ask people in the past what they thought or how they saw the world. But large language models have ingested vast textual corpora produced across time. Could they be used to simulate or approximate historically bounded subjectivities?

Early results suggest that LLMs carry a surprisingly structured representation of historical social structure, one that might be shaped by the cultural record they were trained on. Models perform better for some periods and than others, and their errors are patterned in revealing ways. The project contributes to growing conversations about LLMs as instruments for social science and about the nature of the cultural knowledge embedded in model weights.