Polis is an open-source platform that helps entire cities, states, or even countries find common ground on complex issues.

First launched in 2012, it's been stress tested in tens of thousands of conversations with more than ten million participants worldwide. By collecting and analyzing viewpoints from thousands of participants, Polis reveals points of consensus, even on topics that seem deadlocked.

Polis has become part of the national democratic infrastructure in Taiwan, the UK, and Finland. Taiwan has used it to craft legislation on issues ranging from Uber regulation to revenge porn and online liquor sales; the UK has employed it for national security consultations; and Finland's wellbeing services counties — regional bodies responsible for health and social services — use Polis to design programs like support for elderly safety and mental health services for children, based on what citizens say they need. Governments in Singapore and the Philippines have also adopted the platform, while in Austria, the Klimarat (the National Citizens’ Assembly on Climate) used Polis to bring together thousands of citizens and experts to develop climate proposals.

At the local level, Amsterdam, Bowling Green, Kentucky, and multiple UK cities have used Polis to improve residents' lives. The United Nations Development Programme (UNDP) deployed it for what it called "the largest online deliberative exercises in history," engaging 30,000 youth across Bhutan, East Timor, and Pakistan.

Polis is designed, engineered, and maintained by The Computational Democracy Project (CompDem), a U.S.-based 501(c)(3). The tool has been featured in MIT Technology Review, Wired, The Economist, and The New York Times, and in BBC and PBS documentaries.

Polis 2.0

CompDem is now introducing Polis 2.0, an enhanced version of the original Polis 1.0 platform. This upgraded system combines massive participation capacity — supporting millions of simultaneous participants — with automated mapping of hundreds of thousands of individual viewpoints, real-time LLM-generated summaries, and the ability to keep conversations open indefinitely.

Polis 2.0 achieves this transformative scale through four key mechanisms:

• Scalable Cloud Infrastructure: Polis 2.0’s robust, cloud-powered distributed system scales in real time to meet demand. While Polis 1.0’s largest deployment reached 33,547 participants (a conversation hosted by Germany’s Aufstehen party), Polis 2.0’s infrastructure can handle 10–30x increases, supporting millions of simultaneous participants.
• Dynamic Opinion Mapping: Polis groups participants based on how similarly they vote and what statements they submit. Building on a foundation of simple but solid statistical algorithms from a decade ago, Polis 2.0 now employs more refined and nuanced approaches. These groupings update in real-time as the conversation evolves, maintaining clear analysis even across hundreds of thousands of statements and millions of votes.
• Semantic Topic Clustering: Polis 2.0 is the first to use the Embedding Vector Oriented Clustering (EVōC) library from the Tutte Institute for Mathematics and Computing to automatically organize conversations into evolving topic hierarchies — hundreds of topics and subtopics — drawn from both organizer-seeded comments and participant input. Participants can view all topic areas and select those of greatest interest before entering the discussion. This organic process lets participants collectively shape the agenda over time, with "hot" and "cold" areas of discussion naturally emerging, allowing Polis 2.0 conversations to remain open indefinitely.
• End-to-End Automation: Earlier Polis conversations required intensive moderation of participant input and facilitator expertise to distill dense outputs into actionable reports — processes that demanded extensive training and practice. Polis 2.0 automates conversation seeding, moderation (including toxicity filtering), semantic clustering, and report generation. This removes the expert facilitator bottleneck while preserving the option for human oversight.

How Polis 2.0 works

1. Setting up a Polis 2.0 conversation

Polis is “seeded” with a set of statements that participants can “agree,” “disagree,” or “pass” on.

Polis 2.0 accepts multiple input types. The primary format is short statements (1-3 sentences), optimized for mobile voting — most are generated directly by participants.

Other input formats can be pre-processed using an LLM and entered via CSV:

• Long narratives (chunked into votable statements)
• Workshop transcripts (face-to-face discussion outputs converted to votable statements)
• Social media posts (text from Facebook, Instagram, YouTube, etc., processed and de-duplicated)
• Online media comments (compiled and filtered)
• Email submissions (text-based input from non-digital participants)
• Voice recordings (transcribed and processed into text)

2. Inviting Participants

Polis 2.0 includes multiple systems for managing participant identity and growth:

• Invite Trees: A structured invitation system tracks how participants join conversations, enabling organic growth through networks while maintaining quality. This snowball sampling approach allows organizers to understand how conversations spread and optimize for meaningful participation over viral reach.
• Identity Management: Advanced XID (external identifier) whitelist and download capabilities, plus OIDC authentication providers, ensure secure and flexible participant access.
• Data Portability: Complete data portability with XID support enables cross-platform participant tracking and analysis, compatible with popular polling and survey platforms such as SurveyMonkey, Qualtrics, Typeform, and Google Forms.

3. Participating on Polis 2.0

On Polis 2.0 participants can:

• Select topics of interest — collectively setting the agenda for what everyone will vote on
• Vote on others’ statements — agree, disagree, or pass (there's no reply function, by design)
• Submit statements about issues that matter to them — shaping conversation topics
• Mark which statements are especially important to them (optional)

Multi-lingual capabilities: The system detects a participant’s browser language and automatically translates the UI text and statements into their preferred language. Participants can submit statements in any language and view all statements both in the default language and in their chosen language.

4. Moderating Polis 2.0

Polis conversations with tens of thousands of participant-entered statements require effective moderation. Polis 2.0 includes AI-assisted moderation features to support this:

• Toxicity Detection: Real-time flagging of hate speech, harassment, and extremist content
• Language Processing: Automatic translation for multilingual participation

Human Oversight (recommended):

• Review: Human review of AI moderation decisions
• Cultural Sensitivity: Specialized review for marginalized or under-represented community contributions
• Expert Fact-checking: Specialists verify claims about technical details
• Company and Community Standards: Transparent moderation guidelines co-developed with participant input

In addition, the statement routing system functions as a form of moderation by determining the optimal presentation of statements to each participant.

5. Real-time Analysis & Visualization

Polis 2.0 Outputs

Comprehensive Topic and Opinion Mapping: Polis 2.0 maps the conversation by identifying popular topics, subtopics and their interconnections, areas of consensus, and points of disagreement.

Consensus Statements: For each topic and subtopic, the platform generates collective statements that reflect agreement across all groups, supported by the underlying comments and votes. These statements represent authentic consensus rather than imposed compromise.

Data Repository: All data remains accessible for ongoing reference and further analysis