Understanding the social media echo chamber – Can Mathematics help prevent polarisation online?

How exactly does polarization and populism online work, on a technical level? While the question becomes increasingly important for the cohesion of society, truly answering it remains a huge challenge, especially due to the staggering complexity of relationships and connections on social media. To better understand the dynamics driving those divisive forces, economics professor and Head of the Department of Finance of the University of Luxembourg Christos Koulovatianos is leading the POPULISM-project, which is in large part financed by the Luxembourg National Research Fund (FNR). His team is building mathematical models to quantify the mechanics of populism in online spaces, hoping to provide tools for a society struggling to understand itself. We sat down with him to discuss why an economist is tackling a subject that seems to be more at home in sociology, how you can mathematically model an echo chamber, and whether social media can ever be effectively regulated.

Lëtzebuerger Journal: Your work is on quantifying populism and political polarisation. Most people would associate that with psychology or political science. How come you, as a macroeconomist, are tackling this?

Dr. Christos Koulovatianos: That's a fair question. There's a big gap in the social sciences between practitioners and fundamental research, hiding how mathematical and yet broad fundamental-research economics tools have become. In economics, we learned a hard lesson, especially after the Second World War. Our field was challenged by the hard sciences on whether it was a science at all. We had ideological "civil wars" – Keynesians versus Monetarists, for example – and the only way to find common ground was to mathematise our analysis. When you write down a dynamic system, you can test it. You can see how small changes in parameters can lead to vastly different outcomes – that’s the essence of chaos theory. This quantitative rigor forces clarity and allows for falsifiable statements, which is the bedrock of science.

We believe a similar quantitative rigor is now essential for studying social phenomena like populism. While we have many excellent qualitative descriptions, we lack the tools to model the system's dynamics in a way that people’s decision-making values and beliefs are consistent with these dynamics. We're not trying to be psychologists; we don't pretend to model individual consciousness. Instead, we are approaching human behavior through a "robotics" lens. We build a model of the system and its incentives, and then we see how a rational (or boundedly rational) agent would navigate it. The goal isn't to perfectly describe a person, but to build a useful model of the system that helps us understand its emergent properties.