Scholarlink

<p>The widespread use of aggressive language on Twitter raises concerns about potential negative influences on user behavior. Despite previous research exploring aggression and negativity on the platform, the relationship between consuming aggressive content and users’ aggressive behavior remains underexplored. This study investigates whether exposure to aggressive content on Twitter can lead users to behave more aggressively. Our methodological approach contains four stages: data collection and annotation, aggressive post detection, user aggression intensity metric, and user profiling. We proposed the English Twitter Aggression dataset (TAG-EN) with substantial inter-annotator agreement (Krippendorff’s alpha=0.78). Subsequently, we benchmark the aggression detection performance on TAG-EN dataset (macro F1=0.92) by fine-tuning a pre-trained RoBERTa-large. We quantified user aggression with a proposed “user aggression intensity” metric based on their overall aggressive activity. Our analysis of 14M posts from 63K users revealed that aggressive Twitter feeds can influence users to behave more aggressively online. Furthermore, the study found that users tend to support and encourage aggressive content on social media, which can contribute to the proliferation of aggressive behavior.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

You are what your feeds make you: A study of user aggressive behavior on Twitter

  • 期刊论文

  • 发表日期：

    2025年1月29日

查看全文
  • Swapnil Mane,
  • Suman Kundu,
  • Rajesh Sharma

摘要

The widespread use of aggressive language on Twitter raises concerns about potential negative influences on user behavior. Despite previous research exploring aggression and negativity on the platform, the relationship between consuming aggressive content and users’ aggressive behavior remains underexplored. This study investigates whether exposure to aggressive content on Twitter can lead users to behave more aggressively. Our methodological approach contains four stages: data collection and annotation, aggressive post detection, user aggression intensity metric, and user profiling. We proposed the English Twitter Aggression dataset (TAG-EN) with substantial inter-annotator agreement (Krippendorff’s alpha=0.78). Subsequently, we benchmark the aggression detection performance on TAG-EN dataset (macro F1=0.92) by fine-tuning a pre-trained RoBERTa-large. We quantified user aggression with a proposed “user aggression intensity” metric based on their overall aggressive activity. Our analysis of 14M posts from 63K users revealed that aggressive Twitter feeds can influence users to behave more aggressively online. Furthermore, the study found that users tend to support and encourage aggressive content on social media, which can contribute to the proliferation of aggressive behavior.