Researchers from Google, Harvard, Yale, and Oxford tested a generative-AI system—dubbed the “Habermas Machine”—to broker agreement in politically charged discussions among 5,734 UK participants and a 200-person citizens’ assembly. Participants preferred AI-generated group statements to human-written ones 56% of the time, citing greater clarity and stronger articulation of the majority view, while inclusion of minority views was comparable. AI-facilitated deliberations reduced disagreement; unmediated discussions did not. The findings suggest AI can scale consensus-building while keeping humans in the decision loop, though questions remain over ethical guardrails and responsible deployment in public communications.
Related articles:
