Generative Polis — Consensus-Building with AI, Beyond Polis

As generative AI begins to shape how people make decisions and form values, the question of how AI should be built into the foundations of democracy comes to the fore. Conventional AI alignment has centered on one-directional control — making AI conform to values and norms set in advance by humans. But in a democratic society where diverse values coexist and social norms keep shifting through dialogue, implementing a single "right answer" in AI is not enough. What is needed instead is a framework in which humans and AI influence one another and jointly form shareable meanings and norms.

Taking "Symbiotic Alignment," proposed by the Artificial Life Institute, as its starting point, this session considers how common ground can be found in democratic dialogue in the age of AI. Symbiotic Alignment treats AI not as something to be made to follow human values one-directionally, but as a co-evolving presence that forms social meaning and shared symbols together with people [1]. Grounded in Taniguchi's Collective Predictive Coding (CPC) hypothesis [2], it frames the human–AI relationship as a process in which multiple agents form shared symbols and understanding through interaction.

This perspective carries important implications for large-scale deliberation in digital democracy. Platforms such as Polis [3] have shown the potential to surface opinion groups and points of agreement from participants' voting patterns, revealing common ground across differing positions. Seen through Symbiotic Alignment, such common ground is not merely majority opinion but the shareable values and questions that let people of different positions coexist and keep talking. Beyond that, the "uncommon ground" that Audrey Tang and others have emphasized in the context of Plurality refers to values and questions that people of differing positions in fact share without realizing it — an important clue for finding common ground across coexisting positions, beyond polarization.

Through the theoretical lens of Symbiotic Alignment and an attempt to treat Polis-style large-scale deliberation as a generative model, the session discusses the possibilities for diverse positions to coexist while finding common ground.