Public Spheres for Democracy (PSS4DEMOCRACY)

ERC Consolidator Grant 2025

How can digital news supply strengthen democracy?

Today, algorithms are involved in decisions about which news we see online. However, we barely know which digital public spheres actually help people stay informed, understand politics, and participate in democracy. 

That is precisely what PSS4DEMOCRACY aims to investigate. The project recreates different types of digital spheres in a news app following ideas from democracy research. Next, the project tests digital spaces in Germany, Norway, Poland, and the US with user participation. This will make visible which news offerings best support knowledge about politics, different points of views, and political involvement. 

Within communication and democracy research, there exist manifold theoretical models of democratic public spheres that differ in their ideas on diversity, representation, emotionality or user orientation. Until today, an option to test these models empirically and to systematically compare their actual effects on news usage, political knowledge, and political participation has been missing.  The PSS4DEMOCRACY project pioneers the creation of the technical and methodical conditions for transferring theoretically formulated models of public spheres into algorithmic news systems and subsequently test them in real-life usage situations. Thereby the project paves the way to a substantially new form of empirical research on public spheres. 

Project goals

1. Conceptually developing public spheres
   The project maps prototypical models of public spheres (liberal, deliberative, critical, representative, etc.) and transfers their normative goals, e.g. diversity, inclusion, knowledge gain, into measurable technical criteria for algorithmic systems.
2. Analyzing real-life news systems
   Applying modern NLP techniques, we analyze an extensive, multilingual news corpus from Germany, Norway, Poland, and the US. This enables us to systematically evaluate diversity, biases, visibility of political actors and emotionalization of real-life news services.
3. Developing public-sphere recommenders
   We construct six news recommender models, such as an impartial PSS, a critical PSS, or a knowledge-gain PSS. These use deep pareto reinforcement learning to balance democratic performance goals (e.g. diversity) with user relevance.
4. Testing effectiveness in field experiments
   In one of the most extensive studies on NRS to date, we test the PSS models in a news app in four countries before national elections. We measure, amongst others:
   
   • political knowledge gain
   • affective & ideological polarization
   • perceived diversity
   • involvement & civic spirit
   This is how, for the first time, we gain empirically based knowledge about which news environments strengthen democratic societies.

Project design

The project comprises three work packages:

1. INPUT:
   Systematic literature review, construction of a multilingual news corpus, development of public sphere metrics (PSiNA framework)
2. THROUGHPUT:
   Development of six PSS models, implementation of a news app and NRS pipelines, algorithmic optimization
3. OUTPUT:
   Field experiments in Germany, Norway, Poland, USA; Feldexperimente in DE, NO, PL und USA; derivation of empirically optimal PSS models; creation of guidelines for the design of news recommender systems and policy briefs for media, platform, and AI regulation.

Societal & political relevance

The project creates concrete, technically feasible models of a normatively oriented digital public sphere for the first time. It provides:

• open-source prototypes for news recommenders
• criteria for the evaluation of democratic news environments
• recommendations for the regulation of algorithmic systems
• education and transfer material for media, politics, and the public sphere

PSS4DEMOCRACY thus combines technological innovation with democracy-related research and sets standards for future AI governance in journalism.

Download the project proposal here (PDF).

Framework data