Methods have always been a big part of what I study, particularly longitudinal methods that allow me to take change into account. I do a lot of time-series work, event history analysis, and network analysis, but the common thread is an interest in how social phenomena evolve and change over time. Polarization and misinformation have become central issues in American politics, and I study how they develop and change over time.

One of my recent projects examines polarization in congressional speech. My team looks at one-minute floor speeches in the House. We study who's talking about the same issue and how they're talking about it. From those speeches, we build a network that captures which members are discussing similar topics and use similar language. This allows us to examine rhetorical similarity, as well as how congressional discourse has stayed the same or changed over time.

We’re also studying populism in congressional speech, partly because populism is often viewed as extremely divisive. We want to go back in time to determine whether this is a recent phenomenon that we should all be worried about — or whether we have been through periods of this before. People often point to McCarthyism, for example. If similar periods have occurred before, what helped the country move through them and maintain stable political institutions?

Lastly, I’m doing research on courts, specifically the Supreme Court. There are 70 or so Supreme Court cases heard every term, but the Court is asked to hear approximately 7,000 cases through petitions for certiorari (also known as cert petitions). We are building a database of those petitions from 1946 onward, so that we can study the much larger universe of cases seeking the Court’s attention. Students working on the project are extracting the petition text and then using a large language model to classify the legal and substantive issues. We can then incorporate factors such as participation by interest groups and media attention to examine what affects the issues brought before the Court and the cases it chooses to hear.

Making reliable information available about how courts operate can also improve public understanding of our government institutions. At a time when trust in institutions is very low, presenting clear evidence about how the Supreme Court works is especially important. 

I’ve always been interested in mathematics and analytical problem solving, but I also became fascinated by politics and by questions that affect people's lives. Political methodology brought those interests together. It gave me the opportunity to use quantitative tools not simply for the sake of developing methods, but to answer substantive questions about politics that we couldn't otherwise answer as effectively. That was especially true of my early interest in political processes that unfold over time. What continues to motivate me is that same sense of curiosity: identifying an interesting political question and then figuring out the best way to answer it. What I find especially exciting is that methodological advances can make entirely new questions possible while also providing new answers to long-standing questions.

It’s been really interesting to see the evolution of speech. One somewhat reassuring finding is that, at the level of substantive policy discussion, we see more bipartisan overlap than people might expect. Sometimes the important change is not simply in the rhetoric itself, but in which issues are receiving attention and how they're presented. One of my students looked at environmental speech and how it has changed over time. She found that congressional language about climate change has evolved, including a shift towards terms that might be more politically neutral. 

In my research on populism, we find that populist rhetoric is not confined to presidential politics. It also appears in congressional speech, has been present for a long time, and has been used by both parties. At the same time, there's more overlap and more bipartisanship in discussions of substantive policy issues than you might expect from only looking at the broader populist rhetoric.

Probably the biggest way I've had a public impact is through the databases I create. I put together a database of amicus curiae networks, which captures connections among organizations that work together to file briefs in Supreme Court cases. There’s also the Congressional Blue Slip Database and the Cert Petition Database — everybody seems to be excited about getting their hands on the Cert Petition Database, which is funded by the National Science Foundation. 

I'm trying to make these datasets available for future use by reporters, researchers and students. Most of the students in my American Politics and Methodology (AIM) Lab use data I have collected for their undergraduate thesis work. Creating datasets is a way to make the work extend beyond a journal article. Five, 10 or 20 years from now, someone may use the data to ask a question that we have not even considered yet. That is one of the most rewarding forms of scholarly impact. 

The Agostini Prize was a huge surprise, and I am deeply honored. I see it as a testament to all the students that I have worked with over time, because almost everything I do is coauthored with students, either former students or current students. It was especially meaningful that some former students came back to celebrate at the award ceremony with the Board of Trustees.  I also know having such amazing colleagues around The Oval is one of the reasons I’ve been successful. 

The Certiorari Project is the major focus of my work right now. In addition to creating a data set that will show us how the issues brought before the Court have evolved over time, we want to address the million-dollar question: What makes the Supreme Court more likely to hear a case? We now have a lot of students who are starting to dive in and say well, what's the role of the Solicitor General in this? Or what's the impact of how the brief is written, the quality of the brief? We are also going to look at the moral language used in the briefs that are written, and whether that makes it more or less likely for a case to be taken by the Supreme Court. New large language models can identify rhetorical devices, including those that might be morality-based. We want to see what the evolution of that has been over the last 50 years of the court. 

AI has opened up many new possibilities in the social sciences, because so much of what we do is text-based. The 70 or so Supreme Court cases that are heard every year, we actually know a lot about them because literally, for decades, law students have coded the substantive and legal issues in each. That’s very doable for 70 per year, but it is not feasible to code roughly 7,000 cert petitions per year by hand. This is where large language models become particularly valuable. We can train and validate a model using the cases that legal scholars and law students have already coded. The model can then help us classify the much larger body of cert petitions consistently and at a scale that would otherwise be impossible.  

The conversation above has been edited for brevity and clarity.