Pol.is is an open-source deliberation platform that collects statements from participants and uses unsupervised machine learning to identify groups of opinion and surface consensus statements that cut across those groups. It is associated in AI-governance discussions with large-scale civic consultation, particularly through its use by Taiwan's Digital Ministry.
Mechanism
Participants submit short statements and vote on the statements of others. The platform applies K-means clustering, an unsupervised machine-learning method, to group participants by how they vote, and then identifies consensus statements that transcend group boundaries. Results are visualized by placing participants as points in opinion-space, which can reveal where agreement extends across typical political divides.
Adoption and deployments
Pol.is is developed as an open-source project. It has been adopted by Taiwan's Digital Ministry, the team associated with Audrey Tang, and deployed in civic consultations including the Uber/taxi regulatory debate (2014–2015), vTaiwan processes, and pre-election information-integrity assemblies. It has also been used as one of the deliberation tooling options in Alignment Assemblies contexts.
Relation to AI policy
Pol.is has been cited as a model for large-scale public participation in AI governance, particularly for eliciting preference input at civic scale. The Collective Intelligence Project used Pol.is-mediated deliberation in Taiwan's 2023–2024 AI governance Ideathons.
Relationships
- related: Alignment Assemblies and Collective Constitutional AI — tooling for alignment assembly processes
- related: Deliberative Alignment — broader concept of public deliberation in AI