Which Role Should Robots Play in Democracy? A Bottom-up Perspective

Hans Asenbaum, Emanuela Savini, Maleen Jayasuriya, Janie Busby Grant, Damith Herath, Harshith Ghanta, Friedel Marquardt, Isabella Roberts Abstract: Rapid advances in generative AI and robotics are reshaping democratic practices. AI agents are no longer used solely as tools that support democratic processes, for instance as facilitators or expert systems, but increasingly enter the democratic arena as political agents, including as politicians, activists, or participants. Existing debates on AI and democracy tend to polarize between objectifying AI as a neutral instrument that enables democracy and subjectifying AI by exploring questions of personhood, rights, or citizenship. By contrast, how everyday people envision the role robots should play in democracy remains largely unknown. This paper develops a bottom-up approach to normative theorizing on AI in democracy by drawing on everyday perspectives and participatory design principles. Empirically, it builds on the interdisciplinary project Robots in the Room, which brings together political science, psychology, and robotics. The project conducted six small-group deliberations in which a humanoid robot participated as a peer, articulated political positions, and engaged in argumentation alongside human participants. The analysis draws on 24 qualitative interviews in which participants reflect on their deliberative experiences and theorize the appropriate democratic roles of robots. The findings reveal broad agreement that robots should occupy supporting roles in democratic processes rather than exercising decision-making power, voting rights, or standing as political representatives. Crucially, however, participants do not conceptualize robots in supporting roles merely as technical tools. Instead, robots’ social functions and human-like characteristics are widely perceived as assets that can foster engagement, structure interaction, and enable more authentic and lively democratic deliberation. The paper concludes by discussing the implications of these bottom-up insights for normative democratic theory and the design of AI-mediated democratic institutions.