36 Voices Out of 3,600: What One Percent Participation Teaches About Digital Democracy

Between December 2025 and February 2026, we ran an experiment. We opened a participatory platform — audierne2026.fr — and invited the 3,600 inhabitants of Audierne-Esquibien to co-write a municipal programme. Anonymous forms, public discussions on GitHub, full traceability from submission to integration.

36 citizens contributed. That's roughly 1%.

The instinct is to see this as failure. A thousand hours of engineering for 36 responses. But look closer, and a different story emerges — one about the nature of civic participation, the tools we built too late, and the programme that grew from seeds we didn't expect.

What 36 Contributions Actually Contain

Not all participation is equal. A single contribution about the OPAH-RU housing renewal program came with precise knowledge of ANAH subsidies, downtown building restoration operations, and PLH timelines that no campaign team could have produced alone. A parent's note about school transport for after-school activities revealed a gap no candidate had addressed.

The distribution tells its own story:

Theme	Contributions	Signal
Logement & Urbanisme	11	Housing is the existential concern
Associations & Vie locale	5	The associative fabric wants to be heard
Ecole & Jeunesse	5	Pierre-Le Lec polarizes, but parents mobilize
Economie locale	5	Fishing, tourism, short circuits — identity topics
Environnement	5	Water quality, nesting birds, waste — concrete and local
Culture & Patrimoine	2	Deep knowledge, few voices
Alimentation, Bien-etre & Soins	2	Emerging awareness
Vivre-ensemble	1	Cross-cutting, hard to categorize

The programme page — audierne2026.fr/programme — now carries the weight of these contributions. Every section shows its count. Every engagement can be traced back to a citizen's anonymous form submission, through a contextualized GitHub issue, to a public discussion.

The Tools Behind the Curtain

The platform didn't enrich itself. Behind audierne2026.fr sits a pipeline built within the Ocapistaine project — an AI-powered civic transparency toolkit:

Contribution lifecycle automation:

• Framaforms submissions trigger daily aggregation reports via GitHub Actions
• Each contribution is validated against the Contribution Charter by Forseti 461, our LLM-based charter validation agent, with quality tracked through Opik
• Validated contributions become GitHub issues, enriched with local context and sources
• Mature issues transfer to Discussions for community debate

Programme enrichment pipeline:

• export_contributions_to_md.py extracts all issues and discussions into structured markdown
• prepare_rag_dataset.py builds a searchable knowledge base (documents.jsonl) from docs, PDFs, and contributions
• Category READMEs in docs/<category>/ aggregate local context: OPAH-RU data for housing, Natura 2000 zones for environment, PCAET diagnostics for climate, demographic projections for schools
• Claude-assisted /program-evolution command identifies which contributions are ready to integrate, which discussions to close, and which programme sections need updating

Cross-repo orchestration:

• The audierne2026 repo holds the public site and contribution data
• The ocapistaine repo runs the RAG system, AI agents, and quality monitoring
• The vaettir repo (N8N) handles multi-channel automation
• /sync-to-ocapistaine keeps the knowledge base synchronized across repos

What We're Building This Week

The programme page is live, but it's a foundation, not a finished building. Between now and March 5th, we're using these tools to:

1. Deepen the contextualization — Each of the 7 thematic sections will be enriched with local data, official sources, and cross-references between what citizens asked and what the municipal record shows
2. Map the four lists' positions — The Pierre-Le Lec school debate already shows how a single issue — 4.7M euros, 146 pupils, a debt ratio climbing to 10.8 years — reveals fundamentally different visions for the commune. We're extending this comparative analysis to all themes
3. Publish a RAG-powered comparison tool — Citizens will be able to ask questions about any list's programme and get sourced, traceable answers

Honest Lessons: Why Only 1%

The low participation rate deserves honest examination, not spin. Several factors combined:

What we got wrong:

• Late start — The platform launched in December 2025, barely three months before the election. Participatory culture needs time to take root. Six months minimum, ideally a year
• Form discovery — Framaforms links were buried behind two clicks. A citizen had to find the "Contribuer" page, then select a category, then reach the form. That's two clicks too many for most people
• GitHub as discussion platform — Technically transparent, socially intimidating. Even with a "no GitHub account needed" email fallback, the perception barrier was real
• No physical relay — Digital-only participation in a commune where 51% of the population is over 60 excludes the very people most affected by municipal decisions

What we got right:

• Anonymity — Framaforms collects no personal data. This matters in a commune of 3,600 where everyone knows everyone
• Traceability — Every contribution's journey is auditable, from anonymous form to programme text. No other list can make this claim
• Quality over quantity — The 36 contributions produced actionable, locally-grounded proposals that campaign teams alone would not have generated

An Improvement Playbook for Next Time

If this experiment were to be replicated — in Audierne or elsewhere — here is what we would change:

Improvement	Why	How
Start 12 months before election	Build participatory habits before campaign pressure	Launch platform as a civic observatory, not a campaign tool
One-click contribution	Remove all friction from form access	QR codes on flyers, direct form links, SMS shortcodes
Physical collection points	Meet citizens where they are	Paper forms at the market, pharmacy, post office, church
Regular public restitution	Show that contributions matter	Monthly summary meetings (even 15 min), newsletter
Partner with associations	Leverage existing trust networks	Ask the 250+ Cap Sizun associations to relay the platform
Multilingual and accessible	Include non-native speakers and low-literacy citizens	Simple language versions, audio forms, visual contributions
Real-time feedback loop	Contributors see their impact immediately	Auto-generated "your contribution was received" + "here's what happened"

The Deeper Question

A 1% participation rate on a digital platform in a rural commune of retired fishermen and seasonal residents is not a failure. It's a baseline. The question isn't "why didn't more people use our platform?" — it's "what does meaningful participation look like when you have 3,600 people, four competing lists, and a school renovation that could bankrupt the commune?"

The 36 contributions we received contain more substance than most campaign programmes. The tools we built to process them — from Forseti's charter validation to the RAG-powered knowledge base — can now serve any commune willing to try.

The programme at audierne2026.fr/programme is a living document for five more days. Then it becomes a commitment. We intend to make those five days count.

---

Related: Ocapistaine Kickoff | Charter Validation Testing | Sprint Planning: RAG Prototype