Coordinated online behavior, which spans from beneficial collective actions to harmful manipulation such as disinformation campaigns, has become a key focus in digital ecosystem analysis. Traditional methods often rely on monomodal approaches, focusing on single types of interactions like co-retweets or co-hashtags, or consider multiple modalities independently of each other. However, these approaches may overlook the complex dynamics inherent in multimodal coordination. This study compares different ways of operationalizing multimodal coordinated behavior, examining the trade-off between weakly and strongly integrated models and their ability to capture broad versus tightly aligned coordination patterns. By contrasting monomodal, flattened, and multimodal methods, we evaluate the distinct contributions of each modality and the impact of different integration strategies. Our findings show that while not all modalities provide unique insights, multimodal analysis consistently offers a more informative representation of coordinated behavior, preserving structures that monomodal and flattened approaches often lose. This work enhances the ability to detect and analyze coordinated online behavior, offering new perspectives for safeguarding the integrity of digital platforms.

The rise of online platforms exacerbated the spread of hate speech, demanding scalable and effective detection. However, the accuracy of hate speech detection systems heavily relies on human-labeled data, which is inherently susceptible to biases. While previous work has examined the issue, the interplay between the characteristics of the annotator and those of the target of the hate are still unexplored. We fill this gap by leveraging an extensive dataset with rich socio-demographic information of both annotators and targets, uncovering how human biases manifest in relation to the target's attributes. Our analysis surfaces the presence of widespread biases, which we quantitatively describe and characterize based on their intensity and prevalence, revealing marked differences. Furthermore, we compare human biases with those exhibited by persona-based LLMs. Our findings indicate that while persona-based LLMs do exhibit biases, these differ significantly from those of human annotators. Overall, our work offers new and nuanced results on human biases in hate speech annotations, as well as fresh insights into the design of AI-driven hate speech detection systems.

Proprietary public APIs play a crucial and growing role as research tools among social scientists. Among such APIs, Google's machine learning-based Perspective API is extensively utilized for assessing the toxicity of social media messages, providing both an important resource for researchers and automatic content moderation. However, this paper exposes an important bias in Perspective API concerning German language text. Through an in-depth examination of several datasets, we uncover intrinsic language biases within the multilingual model of Perspective API. We find that the toxicity assessment of German content produces significantly higher toxicity levels than other languages. This finding is robust across various translations, topics, and data sources, and has significant consequences for both research and moderation strategies that rely on Perspective API. For instance, we show that, on average, four times more tweets and users would be moderated when using the German language compared to their English translation. Our findings point to broader risks associated with the widespread use of proprietary APIs within the computational social sciences.

AI-generated counterspeech offers a promising and scalable strategy to curb online toxicity through direct replies that promote civil discourse. However, current counterspeech is one-size-fits-all, lacking adaptation to the moderation context and the users involved. We propose and evaluate multiple strategies for generating tailored counterspeech that is adapted to the moderation context and personalized for the moderated user. We instruct a LLaMA2-13B model to generate counterspeech, experimenting with various configurations based on different contextual information and fine-tuning strategies. We identify the configurations that generate persuasive counterspeech through a combination of quantitative indicators and human evaluations collected via a pre-registered mixed-design crowdsourcing experiment. Results show that contextualized counterspeech can significantly outperform state-of-the-art generic counterspeech in adequacy and persuasiveness, without compromising other characteristics. Our findings also reveal a poor correlation between quantitative indicators and human evaluations, suggesting that these methods assess different aspects and highlighting the need for nuanced evaluation methodologies. The effectiveness of contextualized AI-generated counterspeech and the divergence between human and algorithmic evaluations underscore the importance of increased human-AI collaboration in content moderation.

The aftermath of the 2020 US Presidential Election witnessed an unprecedented attack on the democratic values of the country through the violent insurrection at Capitol Hill on January 6th, 2021. The attack was fueled by the proliferation of conspiracy theories and misleading claims about the integrity of the election pushed by political elites and fringe communities on social media. In this study, we explore the evolution of fringe content and conspiracy theories on Twitter in the seven months leading up to the Capitol attack. We examine the suspicious coordinated activity carried out by users sharing fringe content, finding evidence of common adversarial manipulation techniques ranging from targeted amplification to manufactured consensus. Further, we map out the temporal evolution of, and the relationship between, fringe and conspiracy theories, which eventually coalesced into the rhetoric of a stolen election, with the hashtag #stopthesteal, alongside QAnon-related narratives. Our findings further highlight how social media platforms offer fertile ground for the widespread proliferation of conspiracies during and in the aftermath of major societal events, which can potentially lead to offline coordinated actions and organized violence.

Integrity 2024 is the fifth edition of the Workshop on Integrity in Social Networks and Media, held in conjunction with the ACM Conference on Web Search and Data Mining (WSDM) since the 2020 edition [1-4]. The goal of the workshop is to bring together academic and industry researchers working on integrity, fairness, trust and safety in social networks to discuss the most pressing risks and cutting-edge technologies to reliably measure and mitigate them. The event consists of invited talks from academic experts and industry leaders as well as peer-reviewed papers and posters through an open call-for-papers.

Online social networks are actively involved in removing malicious social bots due to their role in spreading low-quality information. However, most of the existing bot detectors are supervised classifiers incapable of capturing the evolving behavior of sophisticated bots. Here we propose MulBot, an unsupervised bot detector based on multivariate time series (MTS). For the first time, we exploit multidimensional temporal features extracted from user timelines. We manage the multidimensionality with an LSTM autoencoder, which projects the MTS in a suitable latent space. Then, we perform a clustering step on this encoded representation to identify dense groups of very similar users – a known sign of automation. Finally, we perform a binary classification task achieving f1-score =0.99, outperforming state-of-the-art methods (f1-score ≤ 0.97). Not only does MulBot achieve excellent results in the binary classification task, but we also demonstrate its strengths in a novel and practically-relevant task: detecting and separating different botnets. In this multiclass classification task we achieve f1-score =0.96. We conclude by estimating the importance of the different features used in our model and by evaluating MulBot’s capability to generalize to new unseen bots, thus proposing a solution to the generalization deficiencies of supervised bot detectors.

The rise of bots and their influence on social networks is a hot topic that has aroused the interest of many researchers. Despite the efforts to detect social bots, it is still difficult to distinguish them from legitimate users. Here, we propose a simple yet effective semi-supervised method that allows distinguishing between bots and legitimate users with high accuracy. The method learns a joint representation of social connections and interactions between users by leveraging graph-based representation learning. Then, on the proximity graph derived from user embeddings, a sample of bots is used as seeds for a label propagation algorithm. We demonstrate that when the label propagation is done according to pairwise account proximity, our method achieves F 1 = 0.93, whereas other state-of-the-art techniques achieve F 1 ≤ 0.87. By applying our method to a large dataset of retweets, we uncover the presence of different clusters of bots in the network of Twitter interactions. Interestingly, such clusters feature different degrees of integration with legitimate users. By analyzing the interactions produced by the different clusters of bots, our results suggest that a significant group of users was systematically exposed to content produced by bots and to interactions with bots, indicating the presence of a selective exposure phenomenon.

Distance metrics between statistical distributions are widely used as an efficient mean to aggregate/simplify the underlying probabilities, thus enabling high-level analyses. In this paper we investigate the collisions that can arise with such metrics, and a mitigation technique rooted on kernels. In detail, we first show that the existence of colliding functions (so-called iso-curves) is widespread across metrics and families of functions (e.g., gaussians, heavy-tailed). Later, we propose a solution based on kernels for augmenting distance metrics and summary statistics, thus avoiding collisions and highlighting semantically-relevant phenomena. This study is supported by a thorough theoretical evaluation of our solution against a large number of functions and metrics, complemented by a real-world evaluation carried out by applying our solution to an existing problem. Some further research venues are also discussed. The theoretical construction and the achieved results show the soundness, viability, and quality of our proposal that, other being interesting on its own, also paves the way for further research in the highlighted directions.

Microblogs are increasingly exploited for predicting prices and traded volumes of stocks in financial markets. However, it has been demonstrated that much of the content shared in microblogging platforms is created and publicized by bots and spammers. Yet, the presence (or lack thereof) and the impact of fake stock microblogs has never been systematically investigated before. Here, we study 9M tweets related to stocks of the five main financial markets in the US. By comparing tweets with financial data from Google Finance, we highlight important characteristics of Twitter stock microblogs. More importantly, we uncover a malicious practice—referred to as cashtag piggybacking —perpetrated by coordinated groups of bots and likely aimed at promoting low-value stocks by exploiting the popularity of high-value ones. Among the findings of our study is that as much as 71% of the authors of suspicious financial tweets are classified as bots by a state-of-the-art spambot-detection algorithm. Furthermore, 37% of them were suspended by Twitter a few months after our investigation. Our results call for the adoption of spam- and bot-detection techniques in all studies and applications that exploit user-generated content for predicting the stock market.

Recently, DNA-inspired online behavioral modeling and analysis techniques have been proposed and successfully applied to a broad range of tasks. In this paper, we employ a DNA-inspired technique to investigate the fundamental laws that drive the occurrence of similarities among Twitter users. The achieved results are multifold. First, we demonstrate that, despite apparently showing little to no similarities, the online behaviors of Twitter users are far from being uniformly random. Then, we perform a set of simulations to benchmark different behavioral models and to identify the models that better resemble human behaviors in Twitter. Finally, we demonstrate that the number and the extent of behavioral similarities within a group of Twitter users obey a log-normal distribution. Our results shed light on the fundamental properties that drive behaviors of groups of Twitter users, through the lenses of DNA-inspired behavioral modeling techniques. Our datasets are publicly available to the scientific community to further explore analytics of online behaviors.

Recent disasters demonstrated the central role of social media during emergencies thus motivating the exploitation of such data for crisis mapping. We propose a crisis mapping system that addresses limitations of current state-of-the-art approaches by analyzing the textual content of disaster reports from a twofold perspective. A damage detection component employs a SVM classifier to detect mentions of damage among emergency reports. A novel geoparsing technique is proposed and used to perform message geolocation. We report on a case study to show how the information extracted through damage detection and message geolocation can be combined to produce accurate crisis maps. Our crisis maps clearly detect both highly and lightly damaged areas, thus opening up the possibility to prioritize rescue efforts where they are most needed.

Social sensing is based on the idea that communities or groups of people can provide a set of information similar to those obtainable from a sensor network. Emergency management is a candidate field of application for social sensing. In this work we describe the design, implementation and deployment of a decision support system for the detection and the damage assessment of earthquakes in Italy. Our system exploits the messages shared in real-time on Twitter, one of the most popular social networks in the world. Data mining and natural language processing techniques are employed to select meaningful and comprehensive sets of tweets. We then apply a burst detection algorithm in order to promptly identify outbreaking seismic events. Detected events are automatically broadcasted by our system via a dedicated Twitter account and by email notifications. In addition, we mine the content of the messages associated to an event to discover knowledge on its consequences. Finally we compare our results with official data provided by the National Institute of Geophysics and Volcanology (INGV), the authority responsible for monitoring seismic events in Italy. The INGV network detects shaking levels produced by the earthquake, but can only model the damage scenario by using empirical relationships. This scenario can be greatly improved with direct information site by site. Results show that the system has a great ability to detect events of a magnitude in the region of 3.5, with relatively low occurrences of false positives. Earthquake detection mostly occurs within seconds of the event and far earlier than the notifications shared by INGV or by other official channels. Thus, we are able to alert interested parties promptly. Information discovered by our system can be extremely useful to all the government agencies interested in mitigating the impact of earthquakes, as well as the news agencies looking for fresh information to publish.