The echo chamber effect in social networks: theoretical analysis and steering strategies

I present a mathematical framework to model echo chambers in social networks, and steer their impact. Digital communication is ever so pervasive in our societies, as the online and offline worlds become increasingly entangled. In online social platforms, similar-minded users tend to gather, and form strongly opinionated communities. These so-called echo chambers are particularly salient in polarised debates regarding politics, societal issues or conspiracy theories, and tend to foster animosity between opposite sides and fuel reinforcement of pre-existing beliefs. The political and informational landscapes are significantly affected, turning the regulation of echo chambers into a crucial matter of cybersecurity. This calls for a more informed analysis and understanding of this phenomenon. Whether an echo chamber is actually desirable or not is context-specific: my framework is agnostic and able to accommodate both. In the last few years, echo chambers have become a primary focus of research on opinion dynamics and in- formation diffusion, with a plethora of models amenable to shed light on empirical studies. There is however a lack of principled methods to efficiently steer the echo chamber effect. In this PhD thesis I develop two mathematical models to describe, quantify and control the echo chamber effect: a macroscopic one based on group-level dynamics, and a microscopic one incorporating user-level features. For each model, I present algorithms which significantly impact the diversity of content users are exposed to, while accounting for individual preferences to avoid backfire effects. The accuracy of the models and the effectiveness of the recommendation algorithms are illustrated through applications on real-world data. This PhD thesis contributes insights to the benefit of the growing debate on regulation of online social platforms.