Characterizing Twitter Users Who Engage in Adversarial Interactions against Political Candidates

Hua, Yiqing; Нааман, Мор; Ristenpart, Thomas

doi:10.1145/3313831.3376548

Cited by 34 publications

(29 citation statements)

References 44 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Since then, the model has been retrained on a larger dataset and modified to address some of its weaknesses reported by other researchers (e.g., [45]). Several other studies have used the Perspective API and have demonstrated that its predictions are accurate [28,42].…”

Section: Toxicity Annotationsmentioning

confidence: 99%

The Structure of Toxic Conversations on Twitter

Saveski

Roy

2021

Proceedings of the Web Conference 2021

View full text Add to dashboard Cite

Social media platforms promise to enable rich and vibrant conversations online; however, their potential is often hindered by antisocial behaviors. In this paper, we study the relationship between structure and toxicity in conversations on Twitter. We collect 1.18M conversations (58.5M tweets, 4.4M users) prompted by tweets that are posted by or mention major news outlets over one year and candidates who ran in the 2018 US midterm elections over four months. We analyze the conversations at the individual, dyad, and group level. At the individual level, we find that toxicity is spread across many low to moderately toxic users. At the dyad level, we observe that toxic replies are more likely to come from users who do not have any social connection nor share many common friends with the poster. At the group level, we find that toxic conversations tend to have larger, wider, and deeper reply trees, but sparser follow graphs. To test the predictive power of the conversational structure, we consider two prediction tasks. In the first prediction task, we demonstrate that the structural features can be used to predict whether the conversation will become toxic as early as the first ten replies. In the second prediction task, we show that the structural characteristics of the conversation are also predictive of whether the next reply posted by a specific user will be toxic or not. We observe that the structural and linguistic characteristics of the conversations are complementary in both prediction tasks. Our findings inform the design of healthier social media platforms and demonstrate that models based on the structural characteristics of conversations can be used to detect early signs of toxicity and potentially steer conversations in a less toxic direction.

show abstract

Section: Toxicity Annotationsmentioning

confidence: 99%

The Structure of Toxic Conversations on Twitter

Saveski

Roy

2021

Proceedings of the Web Conference 2021

View full text Add to dashboard Cite

show abstract

“…To measure the degree of attack conveyed by tweets toward others on Twitter, we employed Google's Perspective API 11 , a popular tool widely used for online abuse and harassment study [16,17,35]. The perspective API allows users to measure the toxicity of a text in English on a scale from 0 to 1.…”

Section: Measurement Of Toxicitymentioning

confidence: 99%

“…On the other hand, we have to note that there were fewer tweets with very high toxicity overall (e.g., tweets with toxicity scores above 0.7[16,17] or 0.5[18]), regardless of the cluster.…”

mentioning

confidence: 95%

Aggressive behaviour of anti-vaxxers and their toxic replies in English and Japanese

Miyazaki

Uchiba²,

Tanaka

et al. 2022

Preprint

View full text Add to dashboard Cite

The anti-vaccine movement has been gaining traction in many countries since the COVID-19 pandemic began. However, their aggressive behaviour in replies—the form of directed messaging that can be sent beyond follow-follower relationships—is still less understood, and even less so about their differences in languages. We conducted a comparative study of anti-vaxxers’ aggressive behaviours by analysing a longitudinal dataset of COVID-19 tweets in English and Japanese. We found two common features across these languages. First, anti-vaxxers most actively transmit targeted messages or replies to users with different beliefs, especially to neutral accounts, with significantly toxic and negative languages; these replies are often directed to posts about vaccine operations. Second, the influential users with many followers and/or with titles are more likely to receive the most toxic replies from the anti-vaxxers. However, pro-vaccine accounts with a few followers even receive highly toxic replies in English, which calls for a special aid different from the Japanese case. These results provide insights into both language dependent and independent countermeasures against anti-vaxxers’ aggressive behaviour.

show abstract

“…Partisans are also much more likely to share fact-checking messages that denigrate their political opponents and boost their political allies [53]. Highly active partisans are also more likely to engage in adversarial interactions with out-party politicians on Twitter [26,27].…”

Section: Partisanship and Online Behaviormentioning

confidence: 99%

“…Understanding the role of partisanship in social media interactions is integral to improving online platforms. For example, partisanship underscores potentially harmful online behavior such as toxic political discourse and harassment of counter-partisan politicians and members of the public [26,27,34]. Political misinformation on social media has been linked to dire offline consequences such as the January 6 United States Capitol attack [48].…”

Section: Introductionmentioning

confidence: 99%

Birds of a feather don’t fact-check each other: Partisanship and the evaluation of news in Twitter’s Birdwatch crowdsourced fact-checking program

Allen¹,

Martel²,

Rand³

2021

Preprint

View full text Add to dashboard Cite

There is a great deal of interest in the role that partisanship, and cross-party animosity in particular, plays in interactions on social media. Most prior research, however, must infer users’ judgments of others’ posts from engagement data. Here, we leverage data from Birdwatch, Twitter’s crowdsourced fact-checking pilot program, to directly measure judgments of whether other users’ tweets are misleading, and whether other users’ free-text evaluations of third-party tweets are helpful. For both sets of judgments, we find that contextual features – in particular, the partisanship of the users – are far more predictive of judgments than content features. Specifically, users are more likely to write negative evaluations of tweets from counter-partisans; and are more likely to rate evaluations from counter-partisans as unhelpful. Our findings provide clear evidence that users systematically reject content from those with whom they disagree politically. Platform designers must consider the ramifications of partisanship when implementing crowdsourcing programs.

show abstract

Characterizing Twitter Users Who Engage in Adversarial Interactions against Political Candidates

Cited by 34 publications

References 44 publications

The Structure of Toxic Conversations on Twitter

The Structure of Toxic Conversations on Twitter

Aggressive behaviour of anti-vaxxers and their toxic replies in English and Japanese

Birds of a feather don’t fact-check each other: Partisanship and the evaluation of news in Twitter’s Birdwatch crowdsourced fact-checking program

Contact Info

Product

Resources

About