| dc.description.abstract | In this thesis, we study the spread of content related to a conspiracy theory with harmful consequences, a so-called Digital Wildfire (DW). We aim to identify drivers, in complex temporal interaction networks underlying Twitter user activity, to model these phenomena as phase transitions. Furthermore, we investigate the component of The 5G and COVID-19 Misinformation Event that progressed on Twitter in the first half of 2020. The 5G and COVID-19 Misinformation Event is the term adopted for all communications surrounding the alleged connection between the 5G- network and the COVID-19 pandemic and all the real-world implications and consequences that followed. The main goal of the thesis is to lay the foundation for the development of methods that enable us to predict misinformation with the potential of causing harmful consequences. To the best of our knowledge, this thesis is the first attempt at modeling DWs in online social networks (OSNs) as phase transitions. The main finding of the thesis is the identification of characteristics in the dynamics of the communication underlying the DW showing similarities to phase transitions. Furthermore, we identify candidates for the driving forces of the observed transition, namely influential users. The results show a nearly perfect overlap between the vertex with the highest degree centrality and the largest cluster in our network, as well as a minor group of vertices (< 4% of the population) with high degree centrality, at times, being inbound to over half of the edges. These findings suggest that only a few influential users are crucial in driving the conversation on a large scale, i.e., drawing a significant amount of new users to the conversation. Through three community detection algorithms, Leiden [1], Louvain [2], and Label Propagation [3], we can conclude the existence of more than one significantly large conversation cluster. Moreover, the largest conversation cluster at the beginning of the DW does not stay the largest over time. Thus, we find evidence for the DW extending from multiple significant origins. Furthermore, while tracking, we observe oscillations in the largest clusters, where two or more clusters go back and forth between being the largest. Towards the peak on Twitter, we observe an increase in the fraction of the vertices belonging to the top 10% largest clusters, indicating a centralization of the overall discourse. The sum of the observations pointed out in this paragraph indicates that DWs begin from multiple origins of misinformation narratives that more and more become unified towards the peak of the DW. This centralization process is an exciting candidate for an early indicator of misinformation spreading with the potential of becoming a DW. | eng |