From Efficiency to Deliberation: Rethinking AI’s Role in Institutionalizing Democratic Innovations

Authors: Elizabeth Calderón Lüning, Max Stearns, Čedomir Markov Participatory democracy has advanced through deliberative Democratic Innovations (DIs) that foster inclusive citizen engagement, civic education, and reflective deliberation. However, the integration of digital tools and AI-informed platforms into these processes remains underdeveloped. While AI offers the potential to scale participation, reduce costs, and facilitate institutionalization, it also risks prioritizing efficiency at the expense of deliberative depth and democratic trust. In this context, we introduce the concept of “democracy-in-the-loop” (DITL) as a response to the limitations of existing models like “human-in-the-loop” (HITL) and “society-in-the-loop” (SITL). While HITL emphasizes static human oversight and SITL broadens oversight to societal values, both fail to capture the dynamic and iterative nature of democratic deliberation. DITL, by contrast, integrates real-time democratic feedback into AI development and use by embedding “meaningful frictions”—deliberate disruptions designed to provoke critical reflection and foster dialogue. Drawing on findings from four use cases—civil society workshops, policymaker dialogues, and developer engagements conducted in 2024—this study highlights the importance of balancing technical functionality with democratic purpose. Initial findings demonstrate that AI-enabled deliberative systems must act as catalysts for dialogue, not substitutes for human agency. By positioning DIs as “schools of democratic engagement,” we explore how AI can strengthen deliberative processes, making its impacts more effective instead of falling into the tech-solutionist discourse of efficiency at all cost.  Aligning  with the “third wave of AI ethics,” DITL can be an approach not merely fit for AI-informed DIs, but as a framework for integrating AI into systems where democratic oversight is essential.