Modeling communication asymmetry and algorithmic personalization in online social networks

Modeling social interactions and their impact on opinion dynamics has attracted growing interest in recent decades, fuelled by the mounting popularity of online social networks (OSNs). On online social platforms, a few individuals, commonly referred to as \textit{influencers}, produce the majority of content consumed by users. However, classic opinion models do not capture this communication asymmetry in OSNs. We develop an opinion model inspired by observations on leading social media platforms and tailored to the peculiarities of online interactions. Our work has two main objectives: first, to describe the inherent communication asymmetry in OSNs, where a tiny group of \textit{influencers} hegemonizes the landscape of social debate, and second, to model the personalization of content by the social media platform. We derive a Fokker-Planck equation for the temporal evolution of users' opinion distribution and analytically characterize the stationary system behavior. Analytical results, confirmed by Monte Carlo simulations, show how content personalization tends to radicalize user opinion and favor structurally advantaged influencers. These emerging behaviors suggest that algorithmic bias, inherently associated with platform filtering, can lead to undesirable outcomes. As an example application, we apply our model to Facebook during the Italian government crisis in the summer of 2019. Our work provides a flexible framework to assess the impact of algorithmic filtering on the opinion formation process and a fine-grained tool to study the complex interaction between influencers and social network users.