CEU Electronic Theses and Dissertations, 2025
| Author | Andres, Elsa |
|---|---|
| Title | Social contagion mechanisms inference on temporal networks from local and global views |
| Summary | Networks are powerful tools to model systems composed of interacting entities, like societies where individuals are interconnected. These structures are particularly useful for studying the spread of social behaviours such as fashion trends, product adoption, or ideas themselves, as they propagate through human interactions. As societies are dynamic and constantly evolving, temporal networks, where connections between entities change over time, provide an accurate framework to study spreading phenomena. The dynamics of temporal networks can affect propagation processes, thus it is essential to understand their evolution. In particular, their dynamics may evolve on multiple time scales characterising periodic activity patterns or structural changes. The detection of these time scales can be challenging from the direct observation of simple dynamical network properties like the activity of nodes or the density of links. In the first part of the thesis. I propose two new methods based on static representations of temporal networks, which allow us to define dissimilarity metrics and compute their power spectra from their Fourier transforms. I demonstrate these methods outperform the reference measures using synthetic and real-world data sets. One approach identifies more easily periodic changes in network density, while the other one is better suited to detect periodic changes in the group structure of the network. After understanding the characteristics of temporal networks, I delve deeper into the study of contagion processes on networks. The adoption of behaviours is largely determined by stimuli from social interactions or external sources. While some individuals may change behaviour after a single peer’s influence, others require multiple exposures from their social circles or act independently. Those mechanisms, known as simple, complex and spontaneous contagions, often coexist in real-world social contagion processes. The goal of the second part of the thesis is to understand whether coexisting adoption mechanisms can be distinguished at the egocentric network level, without requiring global network information. I formulate this question as a classification problem, employing likelihood analysis and random forest classifiers in various synthetic and data-driven experiments. While this last analysis is conducted on static networks, a more realistic scenario would involve temporal networks where individuals can be infected by both simple and complex contagions. Each person’s behaviour adoption depends on factors like personal characteristics, the propagating behaviour, and the nature of social ties. My objective is to determine which contagion mechanism predominates in social spreading with time-varying interactions. I approach this as a classification problem using a mixed synthetic propagation model on temporal networks. By analysing the simulation curves, I identify three categories of propagation and, through an analytical study, develop methods to detect transitions between them. This study offers a novel perspective on the phenomena occurring at multiple time scales in temporal networks, as well as on the nature of propagation processes. Those insights allow a better understanding of contagion mechanisms from both local and global view, contributing to the broader study of dynamic systems. |
| Supervisor | Karsai, Márton |
| Department | Network Science PhD |
| Full text | https://www.etd.ceu.edu/2025/andres_elsa.pdf |
Visit the CEU Library.
© 2007-2021, Central European University