Socio-Political Analysis and AI Research Projects
Academic Year 2024-2025
Ryan Kennedy. "Utilizing AI-Enabled Moderation to Scale Small-Group Deliberation.”
A large and growing body of research has indicated that experiences deliberating about controversial political issues, especially in a small group context, is highly beneficial to students. Previous research suggest that these experiences are highly satisfying for students, increase voting and political activism, and reduce extremism and affective polarization. With this said, scaling such activities has proven difficult. Organizing these small group activities requires extensive training and personnel, resulting in their limited use. Generative AI holds promise for overcoming this scaling problem by allowing instructors to conduct many small-group online discussion forums simultaneously with minimal intervention. The efficacy and obstacles to achieving this, however, are still unknown. Funds from the GTG initiative will be used to fund two undergraduate researchers to explore the implementation of generative AI moderators. The undergraduate RAs will: (1) learn about deliberation and gain experience as moderators; (2) about generative AI and conduct their own tests of the efficacy of generative AI in particular parts of the moderator role; (3) design prompts and interaction mechanisms for automating parts of the moderator role; (4) conduct experiments using AI agents with these new systems to evaluate their efficacy; and (5) assist in creating a new system for semi-automated moderation.
Sara Watson. "Globalization and the Political Supply Side."
This project is part of a broader research agenda which explores how globalization affects dynamics on the supply-side of the political market—that is, the behavior of candidates and politicians (as opposed to voters). We explore these questions through the lens of France, which has not only witnessed a dramatic increase in trade-induced economic dislocations over the past thirty years, but also rising political fragmentation and important shifts in patterns of inter-party coordination and candidate defection. Our project asks three broad sets of questions. First, what is the relationship between rising economic integration and supply-side political fragmentation, and are there differences across party families? Second, what are the consequences for patterns of ideological polarization? Have candidates from mainstream parties become more polarized over time? Finally, how do mainstream parties respond? Under what conditions are they likely to form alliances against extremist competitors, and under what conditions do such coalitions fall apart? Understanding these dynamics has important implications for emerging work on the domestic consequences of globalization, and the ways in which it can alter representative institutions at the core of democratic resilience. Funds from the GTG initiative will be used to support two undergraduate research assistants for one year to help locate, scrape and analyze individual candidate manifesto data using AI-based text analyses, including structural topic modeling and sentiment analysis. These analyses will allow us to explore questions such as whether the candidates fielded by mainstream political parties have become more varied (or not) in their political ideology and/or rhetoric over time, and how the discourse of coalition versus dissenting coalition candidates diverges.
Academic Year 2023-2024
Janet Box-Steffensmeier. "The Supply Side of the U.S. Supreme Court Docket."
We know a lot about the U.S. Supreme Court cases chosen. We know a lot less about the potential cases, re: the writs of certiorari. This study enhances existing theories of judicial behavior by introducing an exhaustive dataset on writs of certiorari, facilitating novel supply-side inquiries made possible through the application of machine learning into classical themes: 1) how does the supply of cert petitions shape the justices’ agenda-setting behavior; 2) how do external factors, such as interest groups and the media, shape the Court’s docket; and 3) how does the Court’s selection of petitions granted review represent the plurality of public interests?Undergraduate Research Students in Computer Science and Engineering, Data Analytics, and the Social Sciences will work on various parts of the project, including scraping data, building databases, data visualization, model estimation using network, survival and time series analysis, running large language models, as well as research skills such as conducting literature reviews, and writing as a social scientists.
Skyler Cranmer. "Causes and Consequences of Public Figures Facing Military Defeat."
Summary. The last 20 years have seen the U.S. start, fight, and lose two major wars. This project seeks to understand how political elites discussed the stalemate and ends of these conflicts when facing the public. A variety of actor types---e.g., journalists, pundits, special interests, elected officials, military leaders, and more--act and interact to shape the public's view the conflicts. As wars go badly, these different actors tend to produce noticeably different narratives. These diverging narratives tend to polarize and cluster as they compete to dominate the public sphere. The proposed research is a component of a broader research agenda that aims to model these dynamics and the extent to (a) which elites engaged shifted positions and polarized during the ends of the conflicts, (b) responded to, reacted, and led each other, and (c) strategically framed their positions. The literature has yet to understand how unelected public figures shape the public narrative and opinion, particularly during political crises such as failed wars; this broad agenda will help close that gap. (3 undergraduate RAs)Tom Nelson. "Connecting Values to Politics."
Socio-political values like freedom, individualism, and humanitarianism are essential ingredients to public opinion. Individuals support candidates and policy proposals in part because they perceive those political objects as upholding or advancing values that they care about. Unlike attitudes, which vary from positive to negative in valence, values like freedom are appealing to almost all people. The near-universal positivity of values, combined with their centrality to political evaluation, makes values tempting targets for political entrepreneurs determined to shape public opinion in their favor. Political entrepreneurs recruit values by asserting that their policy or program will promote a treasured value, thus burnishing that policy or program in the eyes of the public. In essence, value recruitment amounts to shaping the meaning of values. Investigating the strategic manipulation of value meanings can be accomplished through automated text analysis. The analysis of word embeddings is an exciting new development in text analysis that is perfectly suited to measuring how the meaning of words varies between groups and over time.
As the above examples imply, our current research analyzes values in communication about the environment. Our corpus is the Congressional Record. While this method holds promise, it is clear that this corpus is not suitable for issues that have appeared infrequently in congressional debate. To analyze such issues requires collecting, compiling, and in some cases digitizing, documents from diverse sources such as speeches, press releases, newspaper articles, websites, blogs, newsletters, and others. This is a big task, and we therefore seek support to hire undergraduate assistants to assemble our homegrown corpora.- Ju Yeon Park. "Polarization in U.S. Politics" and "Identifying Interest Groups."
Polarization in U.S. politics has progressed over the past five decades to the point where citizens cannot agree on simple facts. While many studies have examined various institutional factors that contributed to polarization, communicative aspects of polarization have received relatively less attention. This project focuses on the evolution of elite discourse as a potential amplifier of polarization in the public’s understanding of policy issues. Using a cutting-edge AI-based text analysis method, my recent research introduced a new measure, the “grandstanding score,” that captures the intensity of political messages contained in legislators’ public speeches made in congressional committee hearings. Undergraduate Research Assistants are matching witness affiliations to lobbying data, cleaning the congressional committee hearing data, coding speeches, validating the Generative AI's classification of hearings into present topics, and running machine learning models to analyze text data.