The Artificial Citizens’ Assembly: Can AI Deliberate?

What is it about?

Can artificial intelligence debate, disagree, and change its mind like humans do? This study explores whether AI can simulate a citizens’ assembly, where people with different backgrounds and opinions come together to discuss political issues. We developed a structured AI simulation based on theories of deliberative democracy, translating principles such as exchanging arguments, considering different perspectives, and reconsidering opinions into a step-by-step process. We tested it using a Swiss climate referendum, comparing the results with unstructured AI discussions and a real citizens’ assembly. The findings show that AI discussions guided by deliberative theory produce more coherent and higher-quality exchanges, while preserving different viewpoints. However, AI still struggles to reproduce the depth of reasoning and real-life experiences that people bring to democratic discussions. The study suggests that AI cannot replace human deliberation, but it could help researchers and policymakers explore different perspectives and design better democratic processes.

Featured Image

Photo by Nguyen Dang Hoang Nhu on Unsplash

Photo by Nguyen Dang Hoang Nhu on Unsplash

Why is it important?

As artificial intelligence becomes increasingly involved in democratic processes, understanding whether AI can genuinely deliberate, rather than simply produce convincing conversations, is essential. This research shows that how AI discussions are designed matters, and that principles from deliberative democracy can help make them more meaningful, reliable, and transparent. It also offers a way to systematically assess the quality of AI-generated discussions. Ultimately, the study opens new possibilities for using AI to explore political disagreements, test democratic processes, and support better decision-making, while recognising the irreplaceable role of human participation.

Perspectives