Sensemaking at Scale

One of the clearest tests of whether intelligence can be oriented toward coherence rather than fragmentation is the work being done with collective deliberation platforms. The most studied of these is Pol.is, originally developed in response to Taiwan’s Sunflower Movement and now used in policy processes across multiple countries. Pol.is operates on a principle inverse to most social media architectures: rather than amplifying the statements that generate the most engagement, it uses machine learning to map the actual structure of public opinion, identifying clusters of agreement that exist beneath surface disagreement and surfacing the statements that bridge those clusters. Where Twitter rewards outrage, Pol.is rewards what its users call bridging statements—formulations that find resonance across positions that conventional discourse treats as opposed. The same underlying technology that has been used to fracture public conversation is being used here to do the opposite.

Pol.is is one instance of a broader emerging field. Citizens’ assemblies in Ireland, France, and elsewhere have begun integrating AI-assisted transcription and synthesis tools that allow facilitators to track patterns of convergence and divergence across hundreds of hours of deliberation. New Jersey’s State AI Task Force used a pairwise voting platform called All Our Ideas to engage over two thousand workers in prioritizing concerns about AI’s impact on the state’s economy. California is currently deploying an AI-powered deliberation platform to gather community input on recovery from the 2024 Los Angeles wildfires. Researchers at major universities and at organizations like Anthropic and DeepMind have published work on what they call habermas machines—AI mediators that help groups identify mutually acceptable positions rather than driving polarization.

The work is uneven. Some of these deployments are serious; others are theatrical. The Danish artist collective that founded a political party run by an AI model is closer to performance art than to genuine governance innovation. But beneath the noise, a real shift is occurring: the recognition that machine intelligence applied to deliberation, designed against the patterns social media optimizes for, can produce sensemaking outcomes the dominant informational architectures actively prevent. This is what intelligence in service of coherence looks like in practice. Not aspirational. Operational.