From 2018 to 2023, Brazil experienced its most fiercely contested elections in history, resulting in the election of far-right candidate Jair Bolsonaro followed by the left-wing, Lula da Silva. This period was marked by a murder attempt, a coup attempt, the pandemic, and a plethora of conspiracy theories and controversies. This paper analyses 437 million tweets originating from 13 million accounts associated with Brazilian politics during these two presidential election cycles. We focus on accounts' behavioural patterns. We noted a quasi-monotonic escalation in bot engagement, marked by notable surges both during COVID-19 and in the aftermath of the 2022 election. The data revealed a strong correlation between bot engagement and the number of replies during a single day ($r=0.66$, $p<0.01$). Furthermore, we identified a range of suspicious activities, including an unusually high number of accounts being created on the same day, with some days witnessing over 20,000 new accounts and super-prolific accounts generating close to 100,000 tweets. Lastly, we uncovered a sprawling network of accounts sharing Twitter handles, with a select few managing to utilise more than 100 distinct handles. This work can be instrumental in dismantling coordinated campaigns and offer valuable insights for the enhancement of bot detection algorithms. less

Conversations are an integral part of online social media, and gaining insights into these conversations is of significant value for many commercial as well as academic use cases. From a computational perspective, however, analyzing conversation data is complex, and numerous aspects must be considered. Next to the structure of conversations, the discussed content - as well as their dynamics - have to be taken into account. Still, most existing modelling and analysis approaches focus only on one of these aspects and, in particular, lack the capability to investigate the temporal evolution of a conversation. To address these shortcomings, in this work, we present CODY, a content-aware, graph-based framework to study the dynamics of online conversations along multiple dimensions. Its capabilities are extensively demonstrated by conducting three experiments based on a large conversation dataset from the German political Twittersphere. First, the posting activity across the lifetime of conversations is examined. We find that posting activity follows an exponential saturation pattern. Based on this activity model, we develop a volume-based sampling method to study conversation dynamics using temporal network snapshots. In a second experiment, we focus on the evolution of a conversation's structure and leverage a novel metric, the temporal Wiener index, for that. Results indicate that as conversations progress, a conversation's structure tends to be less sprawling and more centered around the original seed post. Furthermore, focusing on the dynamics of content in conversations, the evolution of hashtag usage within conversations is studied. Initially used hashtags do not necessarily keep their dominant prevalence throughout the lifetime of a conversation. Instead, various "hashtag hijacking" scenarios are found. less

This article presents the affordances that Generative Artificial Intelligence can have in disinformation context, one of the major threats to our digitalized society. We present a research framework to generate customized agent-based social networks for disinformation simulations that would enable understanding and evaluation of the phenomena whilst discussing open challenges. less

Regressions trained to predict the future activity of social media users need rich features for accurate predictions. Many advanced models exist to generate such features; however, the time complexities of their computations are often prohibitive when they run on enormous data-sets. Some studies have shown that simple semantic network features can be rich enough to use for regressions without requiring complex computations. We propose a method for using semantic networks as user-level features for machine learning tasks. We conducted an experiment using a semantic network of 1037 Twitter hashtags from a corpus of 3.7 million tweets related to the 2022 French presidential election. A bipartite graph is formed where hashtags are nodes and weighted edges connect the hashtags reflecting the number of Twitter users that interacted with both hashtags. The graph is then transformed into a maximum-spanning tree with the most popular hashtag as its root node to construct a hierarchy amongst the hashtags. We then provide a vector feature for each user based on this tree. To validate the usefulness of our semantic feature we performed a regression experiment to predict the response rate of each user with six emotions like anger, enjoyment, or disgust. Our semantic feature performs well with the regression with most emotions having $R^2$ above 0.5. These results suggest that our semantic feature could be considered for use in further experiments predicting social media response on big data-sets. less

We perform factor analysis on the raw data of the four major neighborhood and shortest paths-based centrality metrics (Degree, Eigenvector, Betweeenness and Closeness) and propose a novel quantitative measure called the Core-Intermediate-Peripheral (CIP) Index to capture the extent with which a node could play the role of a core node (nodes at the center of a network with larger values for any centrality metric) vis-a-vis a peripheral node (nodes that exist at the periphery of a network with lower values for any centrality metric). We conduct factor analysis (varimax-based rotation of the Eigenvectors) on the transpose matrix of the raw centrality metrics dataset, with the node ids as features, under the hypothesis that there are two factors (core and peripheral) that drive the values incurred by the nodes with respect to the centrality metrics. We test our approach on a diverse suite of 12 complex real-world networks. less

Shortly after the first COVID-19 cases became apparent in December 2020, rumors spread on social media suggesting a connection between the virus and the 5G radiation emanating from the recently deployed telecommunications network. In the course of the following weeks, this idea gained increasing popularity, and various alleged explanations for how such a connection manifests emerged. Ultimately, after being amplified by prominent conspiracy theorists, a series of arson attacks on telecommunication equipment follows, concluding with the kidnapping of telecommunication technicians in Peru. In this paper, we study the spread of content related to a conspiracy theory with harmful consequences, a so-called digital wildfire. In particular, we investigate the 5G and COVID-19 misinformation event on Twitter before, during, and after its peak in April and May 2020. For this purpose, we examine the community dynamics in complex temporal interaction networks underlying Twitter user activity. We assess the evolution of such digital wildfires by appropriately defining the temporal dynamics of communication in communities within social networks. We show that, for this specific misinformation event, the number of interactions of the users participating in a digital wildfire, as well as the size of the engaged communities, both follow a power-law distribution. Moreover, our research elucidates the possibility of quantifying the phases of a digital wildfire, as per established literature. We identify one such phase as a critical transition, marked by a shift from sporadic tweets to a global spread event, highlighting the dramatic scaling of misinformation propagation. less

Opinion dynamics is a central subject of computational social science, and various models have been developed to understand the evolution and formulation of opinions. Existing models mainly focus on opinion dynamics on graphs that only capture pairwise interactions between agents. In this paper, we extend the popular Friedkin-Johnsen model for opinion dynamics on graphs to hypergraphs, which describe higher-order interactions occurring frequently on real networks, especially social networks. To achieve this, based on the fact that for linear dynamics the multi-way interactions can be reduced to effective pairwise node interactions, we propose a method to decode the group interactions encoded in hyperedges by undirected edges or directed edges in graphs. We then show that higher-order interactions play an important role in the opinion dynamics, since the overall steady-state expressed opinion and polarization differ greatly from those without group interactions. We also provide an interpretation of the equilibrium expressed opinion from the perspective of the spanning converging forest, based on which we design a fast sampling algorithm to approximately evaluate the overall opinion and opinion polarization on directed weighted graphs. Finally, we conduct experiments on real-world hypergraph datasets, demonstrating the performance of our algorithm. less

Many researchers and organizations, such as WHO and UNICEF, have raised awareness of the dangers of advertisements targeted at children. While most existing laws only regulate ads on television that may reach children, lawmakers have been working on extending regulations to online advertising and, for example, forbid (e.g., the DSA) or restrict (e.g., the COPPA) advertising based on profiling to children. At first sight, ad platforms such as Google seem to protect children by not allowing advertisers to target their ads to users who are less than 18 years old. However, this paper shows that other targeting features can be exploited to reach children. For example, on YouTube, advertisers can target their ads to users watching a particular video through placement-based targeting, a form of contextual targeting. Hence, advertisers can target children by placing their ads in children-focused videos. Through a series of ad experiments, we show that placement-based targeting is possible on children-focused videos and enables marketing to children. In addition, our ad experiments show that advertisers can use targeting based on profiling (e.g., interest, location, behavior) in combination with placement-based advertising on children-focused videos. We discuss the lawfulness of these two practices concerning DSA and COPPA. Finally, we investigate to which extent real-world advertisers are employing placement-based targeting to reach children with ads on YouTube. We propose a measurement methodology consisting of building a Chrome extension to capture ads and instrument six browser profiles to watch children-focused videos. Our results show that 7% of ads that appear in the children-focused videos we test use placement-based targeting. Hence, targeting children with ads on YouTube is not only hypothetically possible but also occurs in practice... less

News frames refer to how journalists organize and present information to convey a particular message or perspective to their readers. When covering elections, these frames shape how the public perceives electoral issues and events. This study examines how news frames, especially conflict and game frames, were employed by news organizations in India to cover the 2014 and 2019 general elections. We analyzed how the frames varied temporally, by region, and by the party being featured in the articles. Key findings include (i) conflict and games frames are employed more often in highly electorally consequential states (higher legislative seats) than in other states (ii) articles featuring challenger parties are more likely to have conflict and game frame articles than those featuring incumbent parties (iii) the national parties (BJP, Bharatiya Janata Party) and (INC, Indian National Congress) disproportionately feature in articles having conflict frames. Overall, our analysis highlights the evolving nature of election campaigns and how conflict and game frames play a major part in them. less

External shocks embody an unexpected and disruptive impact on the regular life of people. This was the case during the COVID-19 outbreak that rapidly led to changes in the typical mobility patterns in urban areas. In response, people reorganised their daily errands throughout space. However, these changes might not have been the same across socioeconomic classes leading to possibile additional detrimental effects on inequality due to the pandemic. In this paper we study the reorganisation of mobility segregation networks due to external shocks and show that the diversity of visited places in terms of locations and socioeconomic status is affected by the enforcement of mobility restriction during pandemic. We use the case of COVID-19 as a natural experiment in several cities to observe not only the effect of external shocks but also its mid-term consequences and residual effects. We build on anonymised and privacy-preserved mobility data in four cities: Bogota, Jakarta, London, and New York. We couple mobility data with socioeconomic information to capture inequalities in mobility among different socioeconomic groups and see how it changes dynamically before, during, and after different lockdown periods. We find that the first lockdowns induced considerable increases in mobility segregation in each city, while loosening mobility restrictions did not necessarily diminished isolation between different socioeconomic groups, as mobility mixing has not recovered fully to its pre-pandemic level even weeks after the interruption of interventions. Our results suggest that a one fits-all policy does not equally affect the way people adjust their mobility, which calls for socioeconomically informed intervention policies in the future. less