How Can AI Help Achieve Effective Deliberation at Scale?

This paper discusses how our research group is using AI, particularly LLMs, to help collectives of humans achieve effective deliberation at scale.1 This paper, in particular: (1) identifies the goals that scalable deliberation should ideally achieve, (2) enumerates the barriers that frequently get in the way of achieving these goals, (3) proposes solutions for overcoming these barriers, informed by our analysis of the relative strengths and weaknesses of humans and LLMs at performing deliberation tasks. The solutions we report in this paper include such concepts as content unbundling, content restatement, creativity cueing, argument structure advising, clustering, participant modeling, and stakeholder analysis. We report current results for implemented solutions when we have them and, in other cases, present ideas that we are in the process of turning into reality. In doing so, we hope to broaden researchers’ images of how they can help achieve effective scalable deliberation at scale, as well as give participants and facilitators a better sense of what the future may hold in terms of how they deliberate in partnership with AI systems.