Search

The massive proliferation of social media data represents a transformative moment in conflict studies. This data can provide unique insights into the spread and use of weaponry, but the scale and types of data are problematic for traditional open-source intelligence. This paper presents preliminary, transdisciplinary work using computer vision to identify specific weapon systems and the insignias of the armed groups using them. There is potential to not only track how weapons are distributed through networks of armed units but also to track which types of weapons are being used by the different types of state and non-state military actors in Ukraine. Such a system could ultimately be used to understand conflicts in real-time, including where humanitarian and medical aid is most needed. We believe that using AI to help automate such processes should be a high-priority goal for our community, with near-term real-world payoffs.

This work advances investigations into the visual media shared by agents in disinformation campaigns by characterizing the images shared by accounts identified by Twitter as being part of such campaigns. Using images shared by US politicians' Twitter accounts as a baseline and training set, we build models for inferring the ideological presentation of accounts using the images they share. Results show that, while our models recover the expected bimodal ideological distribution of US politicians, we find that, on average, four separate influence campaigns -- attributed to Iran, Russia, China, and Venezuela -- all present conservative ideological presentations in the images they share. Given that prior work has shown Twitter accounts used by Russian disinformation agents are ideologically diverse in the text and news they share, these image-oriented findings provide new insights into potential axes of coordination and suggest these accounts may not present consistent ideological positions across modalities.

In this study, we characterize the cross-platform mobilization of YouTube and BitChute videos on Twitter during the 2020 U.S. Election fraud discussions. Specifically, we extend the VoterFraud2020 dataset to describe the prevalence of content supplied by both platforms, the mobilizers of that content, the suppliers of that content, and the content itself. We find that while BitChute videos promoting election fraud claims were linked to and engaged with in the Twitter discussion, they played a relatively small role compared to YouTube videos promoting fraud claims. This core finding points to the continued need for proactive, consistent, and collaborative content moderation solutions rather than the reactive and inconsistent solutions currently being used. Additionally, we find that cross-platform disinformation spread from video platforms was not prominently from bot accounts or political elites, but rather average Twitter users. This finding supports past work arguing that research on disinformation should move beyond a focus on bots and trolls to a focus on participatory disinformation spread., Comment: Published and Peer Reviewed at ACM WebSci 2022

In this paper we present a near-complete dataset of over 3M videos from 61K channels over 2.5 years (June 2019 to December 2021) from the social video hosting platform BitChute, a commonly used alternative to YouTube. Additionally, we include a variety of video-level metadata, including comments, channel descriptions, and views for each video. The MeLa-BitChute dataset can be found at: https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/KRD1VS.

This paper investigates visual media shared by US national politicians on Twitter, how a politician's variety of image types shared reflects their political position, and identifies a hazard in using standard methods for image characterization in this context. While past work has yielded valuable results on politicians' use of imagery in social media, that work has focused primarily on photographic media, which may be insufficient given the variety of visual media shared in such spaces (e.g., infographics, illustrations, or memes). Leveraging multiple popular, pre-trained, deep-learning models to characterize politicians' visuals, this work uses clustering to identify eight types of visual media shared on Twitter, several of which are not photographic in nature. Results show individual politicians share a variety of these types, and the distributions of their imagery across these clusters is correlated with their overall ideological position -- e.g., liberal politicians appear to share a larger proportion of infographic-style images, and conservative politicians appear to share more patriotic imagery. Manual assessment, however, reveals that these image-characterization models often group visually similar images with different semantic meaning into the same clusters, which has implications for how researchers interpret clusters in this space and cluster-based correlations with political ideology. In particular, collapsing semantic meaning in these pre-trained models may drive null findings on certain clusters of images rather than politicians across the ideological spectrum sharing common types of imagery. We end this paper with a set of researcher recommendations to prevent such issues.

In this paper, we characterize the content and discourse on BitChute, a social video-hosting platform. Launched in 2017 as an alternative to YouTube, BitChute joins an ecosystem of alternative, low content moderation platforms, including Gab, Voat, Minds, and 4chan. Uniquely, BitChute is the first of these alternative platforms to focus on video content and is growing in popularity. Our analysis reveals several key characteristics of the platform. We find that only a handful of channels receive any engagement, and almost all of those channels contain conspiracies or hate speech. This high rate of hate speech on the platform as a whole, much of which is anti-Semitic, is particularly concerning. Our results suggest that BitChute has a higher rate of hate speech than Gab but less than 4chan. Lastly, we find that while some BitChute content producers have been banned from other platforms, many maintain profiles on mainstream social media platforms, particularly YouTube. This paper contributes a first look at the content and discourse on BitChute and provides a building block for future research on low content moderation platforms., Comment: This long paper is supplemental to a short version of the paper published in ACM Conference on Hypertext and Social Media 2020

In January 2019, YouTube announced it would exclude potentially harmful content from video recommendations but allow such videos to remain on the platform. While this step intends to reduce YouTube's role in propagating such content, continued availability of these videos in other online spaces makes it unclear whether this compromise actually reduces their spread. To assess this impact, we apply interrupted time series models to measure whether different types of YouTube sharing in Twitter and Reddit changed significantly in the eight months around YouTube's announcement. We evaluate video sharing across three curated sets of potentially harmful, anti-social content: a set of conspiracy videos that have been shown to experience reduced recommendations in YouTube, a larger set of videos posted by conspiracy-oriented channels, and a set of videos posted by alternative influence network (AIN) channels. As a control, we also evaluate effects on video sharing in a dataset of videos from mainstream news channels. Results show conspiracy-labeled and AIN videos that have evidence of YouTube's de-recommendation experience a significant decreasing trend in sharing on both Twitter and Reddit. For videos from conspiracy-oriented channels, however, we see no significant effect in Twitter but find a significant increase in the level of conspiracy-channel sharing in Reddit. For mainstream news sharing, we actually see an increase in trend on both platforms, suggesting YouTube's suppressing particular content types has a targeted effect. This work finds evidence that reducing exposure to anti-social videos within YouTube, without deletion, has potential pro-social, cross-platform effects. At the same time, increases in the level of conspiracy-channel sharing raise concerns about content producers' responses to these changes, and platform transparency is needed to evaluate these effects further.

Information quality in social media is an increasingly important issue, but web-scale data hinders experts' ability to assess and correct much of the inaccurate content, or `fake news,' present in these platforms. This paper develops a method for automating fake news detection on Twitter by learning to predict accuracy assessments in two credibility-focused Twitter datasets: CREDBANK, a crowdsourced dataset of accuracy assessments for events in Twitter, and PHEME, a dataset of potential rumors in Twitter and journalistic assessments of their accuracies. We apply this method to Twitter content sourced from BuzzFeed's fake news dataset and show models trained against crowdsourced workers outperform models based on journalists' assessment and models trained on a pooled dataset of both crowdsourced workers and journalists. All three datasets, aligned into a uniform format, are also publicly available. A feature analysis then identifies features that are most predictive for crowdsourced and journalistic accuracy assessments, results of which are consistent with prior work. We close with a discussion contrasting accuracy and credibility and why models of non-experts outperform models of journalists for fake news detection in Twitter.

This paper introduces LABurst, a general technique for identifying key moments, or moments of high impact, in social media streams without the need for domain-specific information or seed keywords. We leverage machine learning to model temporal patterns around bursts in Twitter's unfiltered public sample stream and build a classifier to identify tokens experiencing these bursts. We show LABurst performs competitively with existing burst detection techniques while simultaneously providing insight into and detection of unanticipated moments. To demonstrate our approach's potential, we compare two baseline event-detection algorithms with our language-agnostic algorithm to detect key moments across three major sporting competitions: 2013 World Series, 2014 Super Bowl, and 2014 World Cup. Our results show LABurst outperforms a time series analysis baseline and is competitive with a domain-specific baseline even though we operate without any domain knowledge. We then go further by transferring LABurst's models learned in the sports domain to the task of identifying earthquakes in Japan and show our method detects large spikes in earthquake-related tokens within two minutes of the actual event.