Polis

The bet

Colin Megill, Christopher Small and Michael Bjorkegren founded Polis in Seattle in 2012, in the aftermath of Occupy Wall Street and the Arab Spring. Those moments had shown both the appetite for mass participation and how incoherent it becomes at scale.

The founders bet that the bottleneck in large-group democracy was legibility. People were willing. If thousands of them could express views in a form a computer could map, an algorithm could surface the consensus hiding inside the noise, and disagreement could become something a group could find its way through instead of something deafening.

Peak

Participants write short statements and vote agree or disagree on one another's. Machine learning sorts them into opinion groups and, most importantly, highlights the statements that win support across those groups. The tool works, and that is no small claim.

Taiwan adopted Polis for its vTaiwan process and used it to deliberate contentious national policy. Amsterdam and other cities ran it, and researchers studied it. In 2018 the founders converted the project into The Computational Democracy Project, a nonprofit that open-sourced the code and looks after the main pol.is deployment.

What worked

Open comment sections never managed to let a crowd disagree productively. Polis does, and it shows where people actually align, without a flame war. The design is elegant. Choosing a nonprofit commons over a proprietary product gave it longevity and trust.

Where a government committed to using its output, as Taiwan did, Polis produced consensus with consequences.

What failed — the mechanism

Polis is a listening instrument, and listening comes before deciding. It can hand a government a rigorously mapped agreement, but nothing in the tool requires anyone to act on it. vTaiwan worked because the state had promised to feed the results into rule-making. Without that promise, the same elegant map is merely interesting.

So Polis's impact is capped by the commitment of whoever deploys it. It stayed a niche instrument because "better consultation" is a smaller thing to sell than participation that also closes the loop. It solves the first half of the problem and leaves the second half, the answer and the consequence, to institutions that may never supply it.

What Agora takes from it

Surfacing what people think is one thing. Connecting it to what government does is another, and Polis shows the gap more cleanly than any other case. Mapping opinion is valuable, but it sits upstream of the mechanic that matters: a dependable response and a consequence. Agora aims to put that loop at the center of the product, though it cannot make an official respond.

Agora also borrows an instinct from Polis. A crowd's signal should be read by group and constituency, not as an undifferentiated mass, and that is what Agora's constituency-weighted analytics are built to do.