Investigating the cross-platform behaviours of online hate groups

Thesis

The past few decades have established how digital technologies and platforms have provided an effective medium for spreading hateful content. Despite efforts from law-enforcement agencies and platform developers to remove or limit such content, online hate ideologies and extremist narratives are still being linked to several catastrophic consequences around the world. The concept of online hate is still considered a complex phenomenon, with its definition evolving across several theoretical paradigms and disciplines, and spanning multiple forms of victimisation. Due to this complexity, research into online hate is fragmented throughout numerous disciplines, including computational social science. Previous research has demonstrated how online hate thrives globally through self-organised, scalable clusters that interconnect to form robust networks spread across multiple social-media platforms, countries, and languages. Although several extensive approaches and methods have been proposed in previous studies for the analysis of online hate, limited research has investigated how hateful behaviours and content compare and relate across different online platforms.

This thesis aimed to address these limitations by developing a cross-platform analysis framework for online-hate researchers to gain a clearer understanding of the dynamics of the global hate ecosystem. More specifically, the designing of this framework involved examining the main functionalities of existing online-hate analysis frameworks, and the extent to which they address cross-platform hate. The strengths and limitations of these approaches then informed the functional requirements of the cross-platform analysis framework. To demonstrate how the framework can provide novel insights into online-hate research, this thesis also details its application to various case studies, including online hate from white-supremacy-supporting users and environments spread during the 2020 US election and the COVID-19 pandemic.

This comprises a comparative analysis of hateful content in terms of the major topics of discussion and psycho-linguistic properties across different types of online platforms using natural language processing techniques. Additionally, the framework is used to explore networks of shared content, particularly through the posting of URLs, by harnessing social-network analysis methods. Finally, the cross-platform analysis framework is validated using a list of validation criteria to evaluate its practicality in investigating hateful content and providing novel insights into the field of online hate. The findings from this can be used to develop more effective analysis tools for online-hate researchers and law-enforcement agencies.

Authors: Zahrah, F

• DOI: 10.5287/ora-zbe4xgqa0
• Type of award: DPhil
• Level of award: Doctoral
• Awarding institution: University of Oxford
• Language: English
• Keywords: cross-platform analysis; computational social science; sentiment analysis; social network analysis; natural language processing; linguistic analysis; social media analysis; online hate
• Subjects: Social media and society; Online hate speech; Computational linguistics--Network analysis