A network science framework for detecting disruptive behaviour on social media

Ashford, James ORCID: https://orcid.org/0000-0002-2678-577X 2022. A network science framework for detecting disruptive behaviour on social media. PhD Thesis, Cardiff University. Item availability restricted.

Preview	PDF - Accepted Post-Print Version Available under License Creative Commons GNU GPL (Software). Download (8MB) | Preview
	PDF (Cardiff University Electronic Publications form) - Supplemental Material Restricted to Repository staff only Download (211kB)

Abstract

Social networking platforms enable individuals to interact with others in a public forum by creating and/or consuming both written and visual content. Due to the popularity and wide-spread adoption of social media, this has led to unforeseen negative consequences where actors use social media to intentionally disrupt normal discourse to subversively influence individuals or groups. As a result, this leads to the problem of detecting anomalous activity, which is challenging due to large quantities of textual information combined with multimedia. Furthermore, this is compounded by issues such as foreign languages. This motivates research into techniques that can detect anomalies in social media activity through language-agnostic approaches. This thesis examines ways in which this can be achieved through network science, using different forms of networks to represent the behaviour of actors in social media, rather than the specific content they have produced. However, diverse affordances on alternative social media platforms make this a complex problem. This thesis examines three alternative classes of network representation with respect to detecting disruption in social media. We examine these representations using techniques from complex network theory. Using a range of social media systems, this thesis provides evidence that network based signals aligning to disruptive behaviours can be detected for alternative forms of social media engagement (e.g., collaboration, message, community and feed-based interactions). Through this approach, this thesis determines prospects for assessing social media in complex and dynamic scenarios without recourse to processing natural language.

Item Type:	Thesis (PhD)
Date Type:	Completion
Status:	Unpublished
Schools:	Computer Science & Informatics
Subjects:	Q Science > QA Mathematics > QA75 Electronic computers. Computer science Q Science > QA Mathematics > QA76 Computer software
Funders:	DAIS-ITA
Date of First Compliant Deposit:	6 July 2023
Last Modified:	06 Jul 2023 13:15
URI:	https://orca.cardiff.ac.uk/id/eprint/160829

Actions (repository staff only)

Edit Item