“Co-construction” in deliberative democracy: lessons from the French Citizens’ Convention for Climate

The paper examines the French Citizens’ Convention for Climate (CCC), a large-scale deliberative mini-public initiated in 2019 to propose socially just measures to reduce greenhouse gas emissions by at least 40% by 2030. Framed as an “innovative procedure for co-construction of solutions,” the CCC juxtaposed the core principle of citizens’ assemblies—reliance on randomly selected citizens—with extensive input from external steering bodies (governance committee, facilitators, experts, legal and technical advisors). The study addresses two questions: whether external input compromised citizens’ agency and responsibility for outcomes, and whether co-construction enhanced resonance with the broader public. Using observations, internal and external surveys, and voting data, the authors analyze agenda-setting, proposal formulation, and decision-making dynamics, and the relationship between the mini-public and the macro-public.

The background situates the CCC at the intersection of deliberative democracy and co-construction. Prior work on mini-publics highlights their potential to improve authenticity, inclusiveness, and consequentiality of democratic decision-making, with mixed evidence on political uptake. Climate issues amplify reliance on expert input due to complexity and technicality and raise challenges for representativeness and trust. Co-construction, while popular in public administration, is conceptually fuzzy and overlaps with co-creation and co-production; it involves citizens and state sharing responsibility and professional support to increase competence and trust. The authors contrast the primacy of citizen-driven deliberation in assemblies with shared accountability under co-construction, raising questions about responsibility for outcomes and the role of non-participants. The CCC provides a testbed for a co-constructive citizens’ assembly in a climate context.

The governance committee accredited up to 40 researchers; 30 social scientists (political science, economics, sociology, philosophy, geography, law) collaborated to collect qualitative and quantitative data over the CCC’s nine-month process. Methods:

• Structured observation: Wide access to plenary and group discussions, in-person and online webinars; researchers wore “Researcher” tags, took notes, and made audio recordings while committing via a charter to non-interference. Attendance never fell below 20 researchers; about two-thirds followed ≥50% of proceedings; five followed 100%. Observers coordinated coverage across parallel sessions and synthesized observations in regular debriefs.
• Internal survey: Repeated questionnaires to CCC participants on values, climate perceptions, attitudes to policies, and views on the Convention. High response in Sessions 1–2 but sharp decline thereafter, then recovery in Session 7. Reported here: Session 1 (N=111–136) on values and climate attitudes; Session 7 (N=63–65) on perceptions of CCC. Instruments and descriptives available online.
• External survey: Two waves of representative samples of the French population (N=1000 each) fielded Apr 22–May 11, 2020 (pre-proposals) and Oct 19–Nov 3, 2020 (post-proposals) to benchmark values/politics, track opinion evolution, and assess perceptions of the CCC using overlapping items with the internal survey.
• Official material and process data: Access to the CCC’s internal platform; use of voting data from the CCC sessions; analysis of the citizens’ final report and legal appendices; review of government follow-up, impact assessments, and parliamentary records.

• External input and citizens’ agency: The governance committee strongly framed agenda and procedures (five themed groups on demand-side sectors; creation and later termination of a cross-cutting ‘squad’), invited experts, and set voting rules. Despite this, citizens preserved agency, notably removing the carbon tax from the agenda after a Session 2 plenary outburst, and reshaping cross-cutting work. Expert and legal support was substantial—unprecedented legal advisory group translated proposals into legal language—yet citizens largely generated ideas that experts refined. Citizens rated external experts (92%) and technical advisors (74%) as key knowledge sources (Session 7, N=65) while perceiving steering bodies more as enablers than originators of measures.
• Decision-making design: Limited training in deliberative methods and late communication of voting rules; block voting likely increased approval rates. Nonetheless, intra-assembly trust was high (mean 7.79/10 for confidence in other groups, N=63).
• Outputs: In Session 7, 149 measures in 43 of 44 blocks were approved; only the 28-hour workweek block was rejected. The motorway speed limit reduction (130→110 km/h) passed with 60% approval, the lowest among adopted blocks. Financing: 75/78 schemes approved (turnout 105), with >70% support for options such as public debt, new taxes (e.g., on wealth/advertising/unhealthy food/carbon border adjustment), and progressive tweaks. Constitutional reforms (Preamble and Article 1 rephrasing) and a referendum on recognizing ecocide were approved; most technical measures were not submitted to referendum by citizens.
• Impacts and follow-up: Government initially claimed support for 146/149 measures but applied “veto cards” and altered others; the omnibus bill’s impact assessment estimated achieving roughly 50–66% of the 40% GHG target, with the High Council on Climate judging the true impact likely lower. Many decrees remained pending; the planned constitutional referendum failed for lack of congressional support.
• Public resonance: External surveys showed low awareness (22% Wave 1; 42% Wave 2) and skepticism about representativeness (74% among CCC-aware respondents in Wave 2) and about government’s sincerity. Yet there was strong majority support (>65%) for measures proposed for referendum. Citizens themselves adopted a distanced stance toward speaking for the broader public (Session 1: 36% speaking for themselves; 19% for the broader public; N=116) and rejected referenda on technical measures by large margins (60–80% against), citing concerns about unenlightened or strategic public voting and anxiety about campaigning.
• Process appraisal by citizens (Session 8): On 0–10 scale, mean grades: government follow-up feeling 3.3 (Q1); adequacy toward 40% target with social justice 2.5 (Q2); CCC contribution to mitigation 6.0 (Q3); potential of citizens’ assemblies to improve democracy 7.6 (Q4).

The co-constructive design provided extensive professional and expert support without eclipsing citizens’ ownership: citizens exercised agenda control and demonstrated independence (e.g., dropping carbon tax; rejecting cross-cutting hierarchy; declining referenda on technical measures). This indicates that significant steering input can coexist with preserved agency, creativity, and freedom of choice. However, the mini-public’s resonance with the macro-public was weak: citizens deprioritized public-facing mechanisms (referendum), and general-public inputs (≈3400 submissions) remained largely unused, while the broader public was skeptical about the process despite supporting many measures in principle. These dynamics led to high internal consensus but limited external consequentiality and political uptake, compounded by an ambiguous “no filter” commitment and executive modifications. The findings address the research questions by showing that co-construction did not dilute responsibility—the citizens remain accountable for outputs—but that it did not, by itself, secure broader public legitimacy or support. A clearer ex ante commitment structure and more systematic public-connection devices (e.g., pairing with referenda or staged public engagement) appear necessary to translate deliberative output into consequential policy change.

The CCC demonstrates that large, expert-supported climate assemblies can produce comprehensive, coherent policy packages and build strong internal consensus, even amid disruptions (strikes, COVID-19). Yet co-construction, as practiced, did not ensure macro-political uptake or public resonance; the government filtered proposals and legal changes stalled, and citizens chose not to leverage referenda on technical measures. Lessons: maintain responsive frameworks that empower citizens; ensure transparent, upfront commitment structures clarifying how proposals will be handled; consider pairing assemblies with referenda to enhance consequentiality and legitimacy; and improve deliberation and voting design (training, clear rules, and more granular voting). The experience suggests citizens’ assemblies can enhance deliberative quality but need robust systemic linkages to deliver broader democratic impact, especially on complex, expert-laden issues like climate policy.

• Research process: Incomplete audio records for many small-group discussions; reliance on researchers’ notes; limited internal survey response after early sessions; one incident of perceived bias from an unpiloted questionnaire item (later corrected); limited facilitation from organizers for researcher-citizen interactions and survey administration.
• Process design: Lack of comprehensive ex ante impact assessment of whether measures collectively achieved the −40% by 2030 target; limited formal training in deliberation and late communication of voting procedures; block voting possibly inflating approval rates; general-public contributions underutilized due to information overload and anonymity.
• Representation and generalizability: Some differences between participants and the public (greater environmental concern and interpersonal trust among CCC participants); public skepticism about representativeness and government sincerity could limit generalizability of attitudinal effects.