Skip to content
Citizen Infrastructure

We the People (2025)

TL;DR. In September 2025, as the United States approached its 250th anniversary, Jigsaw (Google) and the Napolitan Institute ran a national online deliberation called We the People. Over 2,400 people — at least one from every one of the 435 congressional districts — wrote, in their own words, what freedom and equality mean to them, then explored an AI-generated map of the whole conversation and rated draft consensus statements. AI sensemaking distilled more than 1.6 million words into 26 summary statements, most of which drew agreement above 80% across partisan lines. The result is now archived in the American Philosophical Society.

At a glance

Section titled “At a glance”
WhenSeptember 2025
WhereUnited States — online, with participants from all 435 congressional districts
Who2,400+ Americans, recruited for a broadly representative sample. Run by Jigsaw (Google) with the Napolitan Institute
The question”What do freedom and equality mean to you?”
What they didThree phases: open-text answers with AI follow-up questions → an AI-generated interactive opinion landscape of the whole conversation → AI-drafted consensus statements that participants rated
The AIJigsaw’s open-source Sensemaking pipeline distilled 1.6M+ words into 26 summary statements, flagging where people agreed and split
OutcomeMost of the 26 statements reached over 80% agreement across partisan divides; 94% of participants felt personally represented. Archived at the American Philosophical Society’s “These Truths” exhibit

What happened

Section titled “What happened”

Instead of a poll’s tick-boxes or a town hall’s loudest voices, We the People asked open questions and let an AI facilitator follow up on each person’s answer, the way a good interviewer would. That produced more than 1.6 million words — far too much for any team to read. Jigsaw’s Sensemaking tools clustered the responses, filed every statement, and wrote cited summaries, rendering the whole conversation as an interactive opinion landscape rather than a red-versus-blue bar chart. Participants then rated AI-drafted consensus statements, and the 26 that survived mostly cleared 80% agreement across party lines.

The self-reported experience was strikingly positive for a one-hour online exercise: 94% felt their own views were accurately represented, 89% felt it expanded the range of voices they heard, three in four (75%) came away understanding opposing views better, and 97% said they would take part again. Those figures are participants’ own reports, and the exercise was short and entirely online — but the direction is consistent with the paper’s central claim: shown the full distribution of what people actually said, Americans found more common ground than the country’s binary politics implies. Jigsaw open-sourced the conversation and visualisation code so others can reproduce the method.

Why it matters

Section titled “Why it matters”

We the People is the national-scale cousin of Bowling Green’s What Could BG Be?: the same AI sensemaking toolchain, aimed not at recruiting a representative few but at making sense of open-ended input from thousands. Two things make it notable. First, its scale — every congressional district, a genuinely national canvas. Second, its form: by presenting the result as an opinion landscape rather than two opposing bars, it doubles as evidence that how we picture disagreement changes how divided we feel. The honest caveats are the usual ones for this shape of work — self-reported satisfaction, a short online session, and summaries that still need people to interpret them — but as a demonstration that mass, AI-assisted deliberation can surface real cross-partisan agreement, it is one of the clearest yet.

Sources

Section titled “Sources”