Loading...
Information diffusion on polarised networks: empirical analysis and mathematical modelling
An error occurred retrieving the object's statistics
Citations
Altmetric:
Date
2025-12
Abstract
Social media is a phenomenon that is part of the daily life of billions of people around the world. From text posting to sharing news links, social media platforms such as X (formerly Twitter) play an important role in popularising political and controversial topics discussions. However, by showing tailored content to the users, social media platforms potentially deepen their users’ beliefs, which may lead to polarisation, in which users form groups that think similarly while distancing themselves from others that share opposite opinions. Understanding people’s interactions in such polarised environment is of paramount importance as social media has been shaping people’s opinions, interests and way of living.
In this thesis we use social networks tools to analyse Twitter data on discussions of controversial topics, which allow us to unveil polarised structures and to understand how information spreads in such environments. We show that information tends to spread inside polarised groups and rarely spreads between groups.
We model diffusion processes on networks with communities, of which polarised networks are a subset, using multitype branching processes. That allows us to model important characteristics of diffusion dynamics not only on the entire network, but also on each community. Additionally, we derive new quantities specific to diffusion on a network with communities.
We also build a framework to identify the nodes in the network that influence others the most to adopt some content. To do this, we evaluate classical centrality measures extended to temporal networks with communities. We show that we can successfully aggregate nodes into influence bands (a low-score, a mid-score and a high-score band) and can also aggregate centrality scores to analyse the influence of communities over time.
Finally, we introduce the R package CascadeSimulatoR that implements simulations of diffusion processes on networks according to popular models of simple and complex contagion, and summarises the simulation results to calculate statistical characteristics of cascade dynamics.
Supervisor
Description
Peer-reviewed
Publisher
University of Limerick
Citation
ULRR Identifiers
Funding code
Funding Information
Sustainable Development Goals
External Link
License
Attribution-NonCommercial-ShareAlike 4.0 International