The 2nd Computational Political Science for Democracy (CPS4D) working Group seminar futured six presentations of current research projects and broader programs launched in collaboration within and between Kellogg and Lucy institutes.
Michael Coppedge, Professor of Political Science and Faculty Fellow of the Kellogg Institute for International Studies, presented an ambitious project titled AI-driven Modeling of Democratic Development and Decline (AIM-3D). The project enables users to model democratic transitions, simulate scenarios, and forecast regime trajectories. Using V-Dem, GDELT, and democracy-adjacent data (e.g., rule of law, corruption), the team applies advanced AI and machine learning techniques to understand democracy as a complex emergent system. Emphasis was placed on causal discovery, dynamic modeling, and interpretability.
Dmitry Zaytsev, Associate Professor at the Lucy Family Institute for Data & Society, presented his two-decade-long research program titled Computational Social Science Approaches to Modeling Social Change. His work integrates methods from data science, political science, public policy, and sociology to understand how complex social systems evolve over time. The visual roadmap he shared traces the progression of his research from early studies of think tanks and political actors in Russia and the United States to recent investigations into democratization, governance, and polarization. His interdisciplinary approach combines classical multivariate analysis with computational methods (machine learning, social network analysis, natural language processing, and forecasting techniques) to explore the effectiveness of complex systems at national and global levels, bridging questions of policy, business, and society. His work underscores the importance of methodological pluralism and the combination of business analytics tools with political process modeling for real-world impact evaluation and forecasting.
Valya Kuskova, Professor and Associate Director of the Lucy Family Institute for Data & Society, presented the development of her research agenda over the years, showing how it evolved from the application of standard analytical techniques to the integration of theoretical and methodological advancements across business and social sciences, ultimately leading to the creation of novel, data-driven methodologies. Her research trajectory reflects a systematic expansion from psychological and organizational studies to broader applications in business analytics, education, social sciences, and the humanities. Incorporating artificial intelligence, machine learning, and computational approaches such as social network analysis, her work now focuses on predictive analytics and forecasting of complex social problems, impact evaluation in business and education, and the development of advanced network methodologies to address pressing societal challenges such as poverty, inequality, and social cohesion.
Matthew Hauenstein, Assistant Research Professor at the Lucy Family Institute for Data & Society, presented work using natural language processing (NLP) on a dataset of 342 peace agreements to distinguish short-term power-sharing (e.g., patronage unity governments) from long-term consociational arrangements. His findings suggest that short-term models promote internal group cohesion but can hinder democracy by encouraging election delays or authoritarian consolidation – embedding clientelist dynamics in successor parties.
Tom Mustillo, Associate Professor of global affairs in the Keough School of Global Affairs, spoke about a workshop he co-convened at GESIS – Leibniz Institute for the Social Sciences, focused on preparing democracy studies for the transformative impact of AI and digital infrastructure. He emphasized that while democracy research is rich in human-understandable content, it lacks a machine-readable structure. Drawing from FAIR data principles, the workshop aimed to identify necessary steps – like persistent identifiers, ontologies, and formal vocabularies – to build machine-interoperable knowledge systems. The workshop seeks to develop standards and practices that allow democratic research to participate fully in the AI age.
Gessica de Freitas, a Ph.D. student in Political Science and Research Assistant in the Kellogg Institute for International Studies, presented the project “Measuring What Matters: Bill Importance in Multiparty Legislatures.” She introduced a framework to measure bill importance in the Brazilian Congress, moving beyond traditional metrics like bill count. Her multidimensional approach integrates scope, legislative effort, and content complexity, distinguishing symbolic vs. substantive legislation. She discussed text scraping from ~60,000 bills, machine-learning classification (~98% accuracy), and plans to scale expert rankings to the full dataset using predictive models like Random Forests.
