Democratic challenges in an AI-mediated society

Across the world, democracy is under pressure. Political polarisation is deepening, while AI-driven misinformation and algorithmic platforms are changing how people communicate, form opinions, and take part in public life. In this rapidly shifting landscape, established ways of understanding democratic debate are no longer enough.

PICTURE

People influence one another through conversations, media, and political and economic incentives, forming social networks of connected individuals. Over time, these interactions build up into larger patterns – such as political polarisation – which in turn shape how individuals think and act. When algorithms become part of this system, the dynamics change: they can reinforce certain messages, amplify conflicts, and make developments both harder to see and harder to predict.

Research shows that polarisation today is not only about disagreement. It is also about identity and emotion. People may develop strong negative feelings towards those with opposing views, weakening democratic tolerance and, in some cases, increasing support for political violence. These tendencies can be intensified in digital environments, where algorithms favour engaging – and often divisive – content.

At the same time, important questions remain unanswered. We still lack clear ways of understanding how these processes unfold over time, or how different factors – from platform incentives to AI systems – interact to shape political outcomes.

Bringing together political science, economics, engineering, and mathematics, the Theme develops new ways of studying these dynamics. A key ambition is to move towards a kind of “political climate model”: a framework that can simulate how countless small interactions across social networks add up to large societal effects. Representing people as nodes and their influence as the links between them, such models help us understand how political polarisation spreads.

The work centres on three questions:

• How do communication, network interactions, and algorithmic dynamics contribute to political polarisation in contemporary democracies?
• What kinds of interventions or institutional designs could reduce polarisation and support democratic systems?
• And how can we assess the strengths and limitations of mathematical models used to understand society?