How a million people agree

A city of three million can shout, or it can decide. This page is our standing research into the platforms and methods that let very large groups of people find out what they actually agree on — not the loudest voice, not the angriest two percent, but genuine common ground, surfaced and provable. Every entry below was checked live in July 2026, not copied from an old list; the fastest-moving entries were re-checked in August 2026, and where the two checks differ, the later one wins. (Looking for the census of Toronto’s online communities that used to live at this address? It moved to the community census page.)

This is not a dream — it already works

None of the tools below are prototypes. Cities and countries have been deciding with them for a decade:

The consensus platforms, at a glance

PlatformWhat it doesFree / open?Proven where
Pol.is Everyone votes agree/disagree/pass on short statements; the math clusters opinion groups and surfaces the statements that bridge them — consensus made visible. Open source; free to self-host, hosted version free for governments and nonprofits. Taiwan (vTaiwan), and government use from the UK and Finland to Austria’s climate assembly.
Talk to the City Reads thousands of open-ended written answers and produces a clustered, claim-linked report of what people actually said — the open-text cousin of Pol.is. Open source (released 2023); free to self-host. Used with vTaiwan and Chatham House.
Decidim A full participatory-democracy platform: proposals, assemblies, participatory budgeting, citizen initiatives — built for months-long official processes. Open source; free to self-host. Barcelona City Council origin; ~2 million users across processes.
CONSUL Democracy Citizen debates, proposals, voting, participatory budgeting, and collaborative lawmaking. Open source; free to self-host. Madrid origin; 250+ cities.
Your Priorities Idea-crowdsourcing with structured for/against debate — a standing online space where a community keeps talking between gatherings. Open source; free hosted option, genuinely self-serve. Iceland’s government and parliament since 2008.
Consider.it A pro/con slider under a statement, with structured point lists — lightweight deliberation for one question at a time. Open source; the hosted service was in relaunch limbo with no free tier as of August 2026 — self-host only for now. Smaller civic and campus deployments.
Kialo Giant structured argument maps — every claim gets its pro and con tree, so a debate’s whole shape is visible at once. Proprietary, but free for public debates and free for education. Large-scale public and classroom debates.
Ethelo A consensus-optimization engine for trade-offs — participatory budgets and policy packages scored across many criteria at once. Proprietary; managed deployments. Municipalities and health authorities.
Stanford Online Deliberation Platform Video deliberation in small groups of 8–15 with an automated moderator, running many rooms in parallel — built on Deliberative Polling. Hosted by Stanford, access by request only — not self-serve. Deliberative Polling projects worldwide.
Stanford Participatory Budgeting Platform Runs online participatory-budgeting votes: approval, knapsack, and ranked ballots, with honest tabulation. Open source; free. 150+ processes, including Chicago, Boston, Seattle, and New York.
Go Vocal Resident-engagement platform for local governments: surveys, ideation, mapping, participatory budgeting, with AI-assisted sensemaking of the input. Open-core — a free edition is open source; most cities use the paid hosted version. Local governments internationally (formerly CitizenLab).
Metaculus Crowd forecasting — aggregates thousands of people’s probability estimates on real-world questions, a different way of finding what a crowd collectively knows. Public forecasting free; partly open source. Long-running public forecasting community.

There is also an enterprise tier — Remesh, ThoughtExchange, and the big survey suites — that sells AI theme-clustering of large group conversations to institutions at institutional prices. It is real and it works, but nothing in it does anything a community can’t get free from the open tools above, which is why our own work starts there.

Methods, not software

The deepest tools in this field aren’t apps at all — they are ways of convening people, refined over decades:

Three books frame the whole field: Hélène Landemore’s Open Democracy (the case for putting randomly selected citizens at the centre), James Fishkin’s Democracy When the People Are Thinking (the evidence that deliberation changes minds), and Audrey Tang and Glen Weyl’s Plurality (Taiwan’s playbook, free to read online).

Toronto already has the muscle

The most encouraging finding in this whole survey is local. MASS LBP, headquartered in Toronto, has designed and led more than 50 citizens’ assemblies and reference panels across Canada since 2008 — over 300,000 civic-lottery invitations sent, roughly 3,000 volunteer participants — including work for the City of Toronto’s TransformTO climate strategy, Metrolinx, and Toronto Pearson. The craft of convening a representative room of strangers is not something this city would need to import. It lives here.

And the measurement side is arriving on its own schedule: Vox Pop Labs’ Vote Compass, the voter-alignment tool used in CBC-partnered federal elections since 2011, is deploying for 50 Ontario municipalities ahead of the 2026 municipal elections through a Toronto Star partnership — the first municipal rollout at that scale, in this province, right now.

Where AI honestly fits — and where it doesn’t

AI is already inside the best of these tools, doing the part machines do well: the clustering math at Pol.is’s core is machine learning, Talk to the City uses language models to organize thousands of written answers into a readable report, and Taiwan’s civic community has used AI to summarize deliberations since around 2024. Used this way, AI is a microphone for the room — it helps everyone hear what everyone said.

Research goes further. Stanford’s generative-agents work showed AI personas behaving believably in a simulated town, and a 2024 follow-up built “digital twins” of 1,052 real people from two-hour interviews — agents that matched their human’s own survey answers about 85% as consistently as the humans matched themselves two weeks later. Political scientists have shown that language models conditioned on real demographics can reproduce survey response patterns (“silicon sampling”) — and follow-up work shows sharp limits on how representative those simulated voices really are. Canada has a seat in this story too: Advanced Symbolics’ “Polly” forecast public opinion from public social-media signal (CBC profiled it), while pollsters like Abacus Data have publicly argued for real human panels over AI-generated respondents.

Our line, held everywhere on this site: a simulation can rehearse a public conversation — it can help design better questions, stress-test a plan, show what a representative room might look like. It can never substitute for one. Only real people deliberating confer a mandate. Our Assembly Simulator is built and labeled on exactly that line: a rehearsal instrument, never a measurement of what Toronto thinks.

So which should Toronto use?

The full working-through of what’s actually free and self-serve today — including why the honest answer for an online assembly is mostly not a platform at all — is at Choosing tools for bigger rooms, and the city-scale design this research feeds is the One Big Conversation.

← Back to Tribe Map overview