Advancing Deliberation with AI: Architecture, Evaluation, and Simulation

Abstract

This doctoral research revolves around the future of deliberative democracy. In particular I am contributing to the ongoing discussions about designing and assessing scalable deliberative processes and the role of AI in deliberation. Combining deliberative theory, computational social science, and machine learning, my project asks how Large Language Models and adjacent technologies can help extend the democratic and epistemic benefits of deliberation to larger publics without eroding the conditions that make those benefits possible. My cumulative dissertation comprises three interlocking strands: (i) a normative–architectural framework for upscaling deliberation under real-world constraints, (ii) an automated pipeline for measuring deliberative outcomes via the Deliberative Reason Index (DRI), and (iii) an empirical benchmark of LLM agents as deliberators.