Muzhi Liu

I am a Ph.D. student in Political Science at Columbia University, where I am also pursuing an M.A. in Statistics. I work on causal inference in political economy, AI evaluation, and LLM-based measurement.

In political economy, I study how politics and economic life shape each other: how the threat of federal job cuts under DOGE changed government workers’ campaign giving, how America’s global reputation affects demand for American products abroad, and whether a president’s claims of credit for corporate success move political support in the places where those companies employ people.

In my work on AI, I study how AI assistants adapt to what they know about their users, and what that means for who can do good research. One project tests whether an LLM’s research advice becomes stronger, not just more technical, when it remembers that a user has more mathematical training; another models how generative AI lowers some barriers to cross-disciplinary research more than others. I also use LLMs as measurement tools, for example to study how foreign-language publics discuss political violence in the United States.

I continue to work on electoral behavior as well, with projects on Latino voters and on how circadian misalignment affects the cost of voting.

I hold a B.A. in Political Science from Emory University. Beyond my discipline, I am passionate about pure mathematics — especially algebra and analysis — and have taken Ph.D. sequences in analysis, probability, and statistical inference in Columbia’s mathematics and statistics departments.

My research has been supported by the Institute for Humane Studies and APSA’s Centennial Center.