A groundbreaking French Initiative to put AI at the service of democracy
Make.org, Sciences Po and Sorbonne CNRS have joined forces to launch the groundbreaking “Democratic Commons” global research program.
This visionary initiative leverages the immense potential of generative AI to fortify democracy by placing ethics at the heart of AI development.
Democratic Commons stands as a pioneering program for research, development, experimentation, and provision of generative AI solutions as digital commons to enhance democratic processes. The consortium aims to revolutionize citizen engagement by simplifying access to information, offering cutting-edge tools for contributions to complex multilingual democratic debates and by implementing Large Language Models (LLM) tailored to effectively support democracy.
The program's inherent multidisciplinary approach brings together top experts in data science and social sciences. Over a two-year period, more than 50 researchers and engineers will collaborate to develop virtuous applications of generative AI for democracy while simultaneously conducting in-depth studies on the democratic biases of generative AI that could potentially undermine its effectiveness.
In essence, the program will :
➢ Define democratic principles that generative AI must adhere to ➢ Identify biases and errors in LLM that impact democracy ➢ Develop an innovative open-source protocol for evaluating and correcting LLM democratic biases based on this vital conceptual framework
➢ Ensure current tools like Make.org participation platforms use debiased AI at the service of democracy ➢ Ensure future AI generations are inherently compatible with democratic uses
Democratic Commons is endorsed by five world's leading experts in ethical AI: Hugging Face, Aspen Institute, Mozilla.ai, Project Liberty Institute and Genci. A Scientific Board, comprising eminent international figures, will validate the excellence of the results obtained. Committee members include experts from prestigious institutions such as Yochai Benkler (Berkman Klein Center), Hélène Landemore (Yale), Karine Perset (OECD.ai), Raja Chatila (Sorbonne), Djamé Seddah (Inria), Constance de Leusse (AI & Society Institute) and Michelle Barsa (Omidyar Network).
The launch of Democratic Commons was announced by French Secretary of State for the Digital Sector Marina Ferrari during a gathering of France's top AI talents with French President Emmanuel Macron on May 21, 2024.
Democratic Commons is one of seven new winners of Bpifrance's 'Digital Commons for Generative AI' call for projects, conducted as part of the France 2030 investment plan. It will receive financial support of up to 6 million euros from Bpifrance.
This program underscores France's recognized expertise in artificial intelligence on the international stage and continues its historic role in protecting human rights and safeguarding democracy.
Why ?
In an era marked by a profound crisis of confidence in democratic institutions, evidenced by rising electoral abstention, the very foundations of democracy are being questioned. Generative AI, when wielded irresponsibly, has the potential to amplify misinformation and disinformation, empowering those who seek to undermine democratic accountability and erode social trust. Even when used with good intentions, today’s generative AI systems remain poorly understood and riddled with biases that threaten the integrity of our democratic processes.
However, Make.org and its partners firmly believe that AI can be an extraordinary tool for democracy. By enhancing citizen engagement with democratic processes and institutions, AI-powered technologies can foster new relationships among citizens and unlock fresh avenues for understanding and participating in public debates.
It is imperative that democratic authorities, researchers, and practitioners rise to this challenge, creating the conditions necessary for generative AI to contribute positively to sustainable democracy.
The Democratic Commons research program seeks to redefine AI technology as a force for democracy, maximizing its benefits while effectively managing its risks. Make.org, Sciences Po, and Sorbonne-CNRS are pioneering the world’s first initiative dedicated to researching, developing, testing, and deploying generative AI solutions that strengthen democracy.
This initiative embodies what should have been undertaken two decades ago with social networks: leveraging their strengths while ensuring they do not jeopardize democratic values.
The Democratic Commons program directly addresses the urgent need for trusted, transparent, and ethical AI tools within global public institutions.
Soon, thanks to this groundbreaking initiative, states and cities around the world will gain free access to artificial intelligence that honors democratic principles.
Who ?
A multidisciplinary Coalition of the world’s best AI experts
The Democratic Commons project aims to revolutionize AI research by leveraging data from Make.org's participation platforms to identify, measure, and mitigate biases in LLM used on these platforms. Make.org’s platforms will serve as crucial experimental grounds, providing invaluable user-LLM interaction data for cutting-edge research.
The true strength of Democratic Commons lies in the synergy between its components and the complementary expertise of its three founding members. The consortium offers an innovative and experimental research approach in which academic research at SciencesPo and Sorbonne-CNRS will be constantly fed with real data from Make.org platforms. This initiative fosters a dynamic, ongoing dialogue between real-world applications and academic research, creating a powerful feedback loop that drives innovation and ensures practical relevance.
By bridging the gap between theory and practice, Democratic Commons is poised to make significant strides in developing AI systems that are not only technologically advanced but also aligned with democratic values. Through this groundbreaking collaboration, we're not just studying AI's impact on democracy – we're actively shaping a future where technology and democratic principles work in harmony, setting a new global standard for responsible AI development and deployment.
The Democratic Commons Programme will explore several AI capabilities that have the potential to significantly enhance citizen participation:
Summarize political debates
AI will enable citizens at any time to understand the content of, and react to, debates on complex political subjects spanning over several weeks. AI will foster citizen engagement by adapting to less politically-committed citizens.
Writing assistant
AI will accompany each citizen to guide them individually step by step, adapting to their level of knowledge and needs. Tools will enable them to enrich their contribution to a debate in relation to contributions already made by other citizens and ensure contributions from different social backgrounds are equally valued. The program will explore how a generative AI based scientific recommendation system can further consolidate the quality of contributions.
Multilingual translation
Automatic AI translation amplified by generative AI power will enable multilingual debates to take place - meeting the promise of a democratic debate that breaks free of linguistic barriers.
Intelligent Moderation
AI will enable real-time oversight of citizen groups and efficiently distill insights from thousands of contributions, fostering more productive discussions.
These innovations promise to transform how citizens interact with democratic processes, making participation more inclusive, informed, and impactful.
What ?
The Democratic Commons Programme is set to develop the following digital commons:
1. Conceptual Framework Development Establish a conceptual and normative framework to define democratic principles applied to AI uses and identify democratic biases in LLM, utilizing real-world data from Make.org platforms to inform this process.
2. Bias Evaluation Tools Create robust evaluation tools to assess these biases in LLM, grounded in actual usage data, ensuring that the analysis is both relevant and generalizable.
3. Innovative Correction Techniques
Design advanced techniques to reduce hallucinations and biases in generative AI, aligning them with democratic applications.
4. Participatory Democracy Platforms Develop new participatory democracy platforms that leverage bias-corrected generative AI to significantly enhance citizen engagement.
5. Open-Source Solutions
Produce open-source LLM and participation platforms that are corrected for democratic biases, facilitating their use in electoral, executive, and legislative processes.
The resulting conceptual framework and evaluation tools will be made available as open-source resources, inspiring the AI community to create “safe for democracy” models. This initiative aims to empower the integration of AI into citizen participation systems, with Make.org's platforms slated for open-source release once they are secured and technically stabilized. Together, these efforts promise to advance the role of AI in fostering democratic engagement and participation.
-1.png)
.png)
