Dynamics of information spread in online social systems: a Twitter-based study

The spread of misinformation on social media can have profound social and political implications; therefore, understanding what features contribute to its virality is essential. This thesis investigates the dynamics of information diffusion on Twitter (now X), with a focus on both true and false fact-checked claims across multiple languages and topics. Retweet data are used to reconstruct diffusion cascades, while follower relations provide a realistic network structure. The empirical analysis reveals that cascade size distributions are heavy-tailed, and more importantly, demonstrates that true and false cascade sizes follow the same distribution, in line with recent findings on misinformation diffusion. Further, this thesis explores various diffusion processes on the Twitter followers network. While standard diffusion models (Independent Cascade, SIR), which have a constant transmission probability, fail to reproduce empirical results, this thesis find that better predictors are models where the spreading probability varies from node to node, i.e., it depends on the ratio of the sender's and receiver's degrees. Finally, different mitigation strategies are evaluated to identify which users are most suitable for targeted interventions (such as providing information about content veracity or restricting the visibility of specific posts) with the aim of reducing the spread of misinformation. The study highlights the importance of network structure and node heterogeneity in shaping the spread of online information and offers insights for countering misinformation.