AI Consolidates Discussion of 250 People into One

In commemoration of the 250th anniversary of the founding of the United States, an experiment in large-scale group deliberation using AI was conducted. Using the 'Thinkscape' platform developed by AI startup Unanimous AI, 277 Americans randomly selected from across the nation participated in an approximately 20-minute online deliberation. They discussed the question "What are the top 3 innovations America has contributed to the world over the past 250 years?" and derived a collective answer.

Large-group discussions have an inherent difficulty in succeeding. In typical business meetings or focus groups (small groups for gathering opinions), the quality of discussion deteriorates once participants exceed 8 to 10 people. This is because there is insufficient time for everyone to speak, and "coherent conversations" involving hundreds of people have been impractical until now. This experiment directly addressed this fundamental challenge with technology.

Unanimous AI adopted what is called 'hypercommunication' technology. The approach divides participants into multiple small groups that conduct parallel discussions, while specialized AI agents connect all small groups in real-time, allowing the entire process to function as a single deliberation. In this experiment, 277 people were divided into parallel deliberation spaces of 4-5 people each, with AI agents integrating these discussions.

During the deliberation process, participants could express their opinions through text, voice, or video. A total of 94 ideas were submitted by the group as a whole, followed by a filtering process. The participants had diverse political and social backgrounds and were recruited broadly from different regions across the nation. A distinctive feature was that it was designed as 'deliberation' involving substantiation and counterargument, rather than simple survey responses.

The background to this experiment's prominence is growing interest in 'collective intelligence' through AI. The concept is that rather than simply adding together the knowledge or judgments held by many people, one can achieve higher-quality collective decision-making by refining them through discussion. Unanimous AI refers to this concept as 'collective super-intelligence,' and it has been covered by media outlets like VentureBeat since about two years ago.

These efforts to realize large-scale deliberation through AI are seen as having potential applications across a wide range of domains, including democratic decision-making, policymaking, and consensus-building within organizations. However, the transparency of how AI agents adjust the flow of discussion and the question of how faithfully participants' opinions are reflected will likely become focal points in future discussion. As a proof-of-concept demonstrating the viability of the technology, this experiment represents a significant first step.