AIM-3D Team Presents Research at APSA 2026
The AIM-3D team participated in the 2026 American Political Science Association (APSA) Annual Meeting, held September 3–6, 2026, in Boston, Massachusetts.
Dmitry Zaytsev presented the paper “AIM-3D: A Prototype Neural Network for Modeling Democratization,” co-authored with Michael Coppedge and Valentina Kuskova, as part of the panel “Democracy and Democratic Erosion in the Age of Artificial Intelligence.”
The paper introduces a new methodological pipeline for AI-based modeling of complex social processes, designed to move beyond key limitations of conventional linear path models. The multi-stage workflow combines neural-network-based causal scoring, forecast-necessity testing, estimation of signed net causal effects, and characterization of nonlinear functional forms.
Applied to long-term democratic development, the pipeline makes it possible to distinguish high-scoring relationships from those that are actually necessary for forecasting, identify feedback loops, and uncover nonlinear dynamics that conventional linear models may not capture.
This research contributes to AIM-3D’s broader effort to develop AI-driven methodological tools for modeling, forecasting, and understanding complex social and political processes.
Authors: Michael Coppedge, Dmitry Zaytsev, and Valentina Kuskova
Presented by: Dmitry Zaytsev
Conference: APSA 2026, September 3–6, 2026, Boston, Massachusetts
Paper: AIM-3D: A Prototype Neural Network for Modeling Democratization
Read more in our Publications
