Updates

A Year On: How the Democratic Commons is shaping the future of AI and Democracy

Updates

A Year On: How the Democratic Commons is shaping the future of AI and Democracy

Share this article

In September 2024, Make.org, Sciences Po, Sorbonne and CNRS joined forces to launch the groundbreaking “AI For Democracy Democratic Commons” global research program. For over two years, more than 50 researchers and engineers are exploring how generative AI can be aligned with democratic values, while fighting against biases, disinformation, and exclusion. Funded as one of Bpifrance’s “Digital Commons for Generative AI” winners, the programme aims to build open, ethical, and inclusive AI tools to strengthen citizen participation and democratic dialogue.

A year in, our Progress Report, accessible below, provides an overview of the key results and outputs accomplished since the Programme’s launch.

Democracy in Crisis

Declining trust in institutions, rising polarisation, and the spread of disinformation are symptoms of deeper democratic stress. Many citizens feel excluded, unheard, or overwhelmed by recent events.

Generative AI systems add a new dimension: they have potential to amplify misinformation or propagate subtle biases, or reproduce opaque decision-making. At the same time, if used responsibly, they can lower cognitive and linguistic barriers, help more people engage meaningfully, and distill insight from large scale debates.

That’s why Make.org and its partners questioned if AI can be built, used, and governed in a way that serves democracy rather than threatening it.

In order to answer that question, Make.org, Sciences Po, Sorbonne and CNRS decided to create this programme, combining technical, social-scientific, and civic expertise:

  • Make.org leads the project, providing experience in large-scale civic participation platforms and expertise in applying AI in that context.
  • Sciences Po (Cevipof and Medialab) contributes with theoretical and empirical research on democratic principles, deliberation, and AI biases.
  • Sorbonne University and CNRS develop evaluation and correction protocols to detect and mitigate democratic biases in AI systems.

The consortium is supported by partners such as Hugging Face, Mozilla.ai, Aspen Institute, Project Liberty, Genci, the AI & Society Institute, and the Center for AI and Digital Policy. A Scientific Supervisory Board of leading international experts ensures ethical and scientific rigor.

What Democratic Commons is building

The Democratic Commons programme advances through a series of interdisciplinary workstreams, bridging political theory, computational modeling, and civic innovation. Here are the key developments achieved between March and August 2025:

  • Dialogue AI: a new citizen platform. Make.org has launched Dialogue, a multilingual, modular platform for citizen co-construction. It reimagines participation as a sequence of interactive views where users can vote, comment, and collaborate. The first international use case will debut at the Youth Assembly Forums 2025 in Canada, enabling young people to co-create recommendations on the societal impact of AI.
  • Defining democratic principles for AI: CEVIPOF at Sciences Po has developed a refined platform highlighting five core democratic principles (participation, pluralism, deliberation, responsibility, and agreement/disagreement), complemented by four cross-cutting values (transparency, agency, autonomy, and equality). This framework is now guiding all teams in identifying and mitigating “democratic biases”.
  • Characterising democratic biases in AI: Sciences Po’s Medialab is mapping how algorithms shape citizen participation. From analysing deliberative platforms such as Remesh and Pol.is, to integrating social choice theory metrics, their work helps uncover where AI may amplify or reduce democratic diversity.
  • Evaluating and correcting biases: Sorbonne and CNRS have created evaluation protocols to measure how LLMs adhere to democratic principles across activities like translation, summarisation, and writing assistance. They are also developing correction tools, including multilingual representation models, argument clarification corpora, and stance-preserving translation metrics, all aimed at reducing democratic distortions in AI systems.

Together, these efforts are laying the foundations for open-source democratic AI:

  • Social science publications defining democratic principles for AI.
  • Bias evaluation and correction tools.
  • Open-source, bias-corrected LLMs.
  • Three Make.org participation platforms powered by these models.

Sharing and growing the Commons

The programme’s dissemination strategy is scaling fast. A dedicated LinkedIn page and newsletter (already reaching over 300 experts and practitioners) share updates, publications, and calls for collaboration.

In 2025 alone, the project was showcased at over 25 international events, including the European Parliament, the University of Oxford, and the International Joint Conference on Artificial Intelligence (IJCAI).

It was also shortlisted for the Council of Europe’s 2025 Democracy Innovation Award, selected from 500 global initiatives.

Towards AI for Democracy

Democratic Commons is more than a research  project, it is a collective effort to ensure that the technologies shaping our future reinforce democracy rather than eroding it. By combining innovation, academic rigor, and citizen participation, this program is building open tools that any government, institution, or civic actor can use to enhance public deliberation.

At Make.org, we believe democracy thrives when citizens are both empowered and protected. With Democratic Commons, we are taking a decisive step to ensure AI becomes an ally of democracy, not its adversary.

👉 Learn more about the Coalition here: Democratic Commons.
📩Get in touch with us: communs-democratiques@make.org