France's Great National Debate (2019)
TL;DR. After the Yellow Vests protests, President Macron launched the Grand Débat National in early 2019: a two-month, nationwide invitation for the French to weigh in on four big questions. Most of it was “deliberation in the wild” — comment boxes and self-selected town halls — and little of it changed policy. But buried inside were 21 randomly-selected regional assemblies, and those converged on a clear idea: create a new democratic body for the ecological transition. That convergence convinced Macron to launch the Citizens’ Convention for Climate.
At a glance
Section titled “At a glance”| When | January–March 2019 (two months) |
| Where | France, nationwide |
| Trigger | The Yellow Vests (gilets jaunes) movement, sparked by a fuel/carbon tax (Nov 2018) |
| Reach | ~1.5% of the population engaged directly; 2.8M online visitors; ~500,000 in local meetings; ~1,400 in regional assemblies |
| Selection | A mix: self-selected (online + town halls) plus 21 regional assemblies by stratified random sampling |
| Topics | Four set questions: taxation, the organisation of the state, democracy & participation, the environment (immigration was kept off the agenda) |
| Outcome | Seeded the Citizens’ Convention for Climate (2019–2020) |
What happened
Section titled “What happened”The Yellow Vests, who began protesting a fuel-tax rise in November 2018, became the longest French protest movement since 1945. Facing a choice between repression and dialogue, Macron chose dialogue, writing an open letter inviting citizens into a debate on four themes “to build together a republic of permanent deliberation.” For two months, ordinary politics was effectively paused and a sprawling ecology of formats ran at once:
- a government online platform (2.8M visitors) that only allowed comments — no deliberative features;
- local meetings of self-selected participants (~500,000 people);
- 21 regional citizens’ assemblies, chosen by stratified random sampling (~1,400 people, two days of facilitated deliberation each);
- grievance books in town halls and schools (an echo of 1789), emails, terminals in train stations and post offices to reach the hard-to-reach, and national thematic conferences with unions and associations.
Inputs to outputs
Section titled “Inputs to outputs”The roughly 6 million online contributions were processed by automatic text analysis from two private firms; everything else — emails, grievance books, the regional assemblies’ written output — was synthesised using older techniques such as the “knowledge trees” of Michel Serres. Because the overall design was haphazard, there was no clear way to rank these very different inputs against each other, and little of the contribution ended up shaping policy.
Impact
Section titled “Impact”The one decisive result came from the randomly-selected assemblies: a dozen of them independently converged on the idea of new democratic governance for the ecological transition. That convergence — the sense that if many random samples of the public land in the same place, the wider public probably would too — is widely credited with persuading Macron to launch the Citizens’ Convention for Climate later that year, a far more structured and consequential citizens’ assembly.
How it went
Section titled “How it went”By the political theorist Hélène Landemore’s reading (she attended one of the regional assemblies in Rouen), the Great National Debate was mass participation but not mass deliberation. The self-selected meetings drew mostly older, time-rich Macron supporters; with no facilitation, the “usual suspects” dominated and the Yellow Vests themselves were largely absent. The randomly-selected assemblies were the bright spot — randomness reached the people self-selection misses, and the yellow vests were in the room. It was also, Landemore notes, a low-tech exercise where AI could have helped — as a facilitator distributing speaking time, a fact-checker, or a synthesiser — but wasn’t used. The lesson points toward scaling deliberation by quality, not just headcount, and toward AI reflectors that help a group understand itself.
Sources
Section titled “Sources”- Hélène Landemore, “Can AI Bring Deliberation to the Masses?” — Stanford HAI seminar, 2025: youtube.com/watch?v=m_EkUpH5L4Y
- “Regional Citizens’ Conferences at the time of the French Grand Débat National” — Missions Publiques: missionspubliques.org
- “The 2019 Grand Débat National in France: A Participatory Experiment of Limited Legitimacy” — Democracy International: democracy-international.org