Advanced search
1 file | 18.29 MB Add to list

Identifying online misinformation and disinformation spreading using graph theory

Bart De Clerck (UGent)
(2024)
Author
Promoter
(UGent) and Filip Van Utterbeeck
Organization
Abstract
In the digital age, social media platforms have transcended their original purpose of connecting people, mor- phing into powerful tools for influencing public opinion and shaping political landscapes. As these platforms increasingly serve as primary sources of news and information, they also become battlegrounds for various actors — from political entities to clandestine groups — seeking to manipulate societal beliefs and behaviors. This manipulation, often subtle and insidious, leverages the complex interplay of algorithms, echo chambers, and network effects to spread misleading information at scale. The consequences of such activities can be pro- found, affecting not only individual perceptions but also public trust and the integrity of democratic processes. This thesis investigates the manipulation of public opinion on social media platforms through the lens of complex systems theory. Central to our approach is the development of network-based representations that model social media interactions as complex networks. We explore various network representations, focusing particularly on the use of maximum entropy models, to identify statistically significant interactions and filter out noise within these networks. The research also integrates Natural Language Processing (NLP) techniques and temporal analysis to uncover more subtle forms of coordinated behavior and information diffusion. We provide a broader context for the research by examining the historical evolution of information ma- nipulation, illustrating the shift from traditional media to digital platforms and the corresponding changes in manipulation tactics, and how these can be detected and countered. Next, we explore the concepts of graph randomization and maximum entropy models, leading to the creation of MaxEntropyGraphs.jl, a Julia package specifically designed for efficient graph randomization and analysis. This tool offers several benefits, includ- ing high performance and versatility, making it suitable for a wide range of applications beyond social media network analysis. The effectiveness of the methodologies is demonstrated through the analysis of real-world disinformation campaigns, including an analysis of datasets from the Twitter information operations report, and a detailed study of a Belgian disinformation case. These studies not only validate our methodologies but also reveal in- sightful patterns and tactics used in modern disinformation campaigns. The thesis concludes by underscoring the increasing challenges posed by the evolving landscape of information manipulation, particularly with the rise of generative AI, and highlights the need for interdisciplinary collaboration and continuous adaptation of detection strategies.

Downloads

  • (...).pdf
    • full text (Published version)
    • |
    • UGent only
    • |
    • PDF
    • |
    • 18.29 MB

Citation

Please use this url to cite or link to this publication:

MLA
De Clerck, Bart. Identifying Online Misinformation and Disinformation Spreading Using Graph Theory. Ghent University. Faculty of Economics and Business Administration, 2024.
APA
De Clerck, B. (2024). Identifying online misinformation and disinformation spreading using graph theory. Ghent University. Faculty of Economics and Business Administration, Ghent, Belgium.
Chicago author-date
De Clerck, Bart. 2024. “Identifying Online Misinformation and Disinformation Spreading Using Graph Theory.” Ghent, Belgium: Ghent University. Faculty of Economics and Business Administration.
Chicago author-date (all authors)
De Clerck, Bart. 2024. “Identifying Online Misinformation and Disinformation Spreading Using Graph Theory.” Ghent, Belgium: Ghent University. Faculty of Economics and Business Administration.
Vancouver
1.
De Clerck B. Identifying online misinformation and disinformation spreading using graph theory. [Ghent, Belgium]: Ghent University. Faculty of Economics and Business Administration; 2024.
IEEE
[1]
B. De Clerck, “Identifying online misinformation and disinformation spreading using graph theory,” Ghent University. Faculty of Economics and Business Administration, Ghent, Belgium, 2024.
@phdthesis{01J8F1HE8KPDXZNNKJYVK08S2W,
  abstract     = {{In the digital age, social media platforms have transcended their original purpose of connecting people, mor- phing into powerful tools for influencing public opinion and shaping political landscapes. As these platforms increasingly serve as primary sources of news and information, they also become battlegrounds for various actors — from political entities to clandestine groups — seeking to manipulate societal beliefs and behaviors. This manipulation, often subtle and insidious, leverages the complex interplay of algorithms, echo chambers, and network effects to spread misleading information at scale. The consequences of such activities can be pro- found, affecting not only individual perceptions but also public trust and the integrity of democratic processes.

This thesis investigates the manipulation of public opinion on social media platforms through the lens of complex systems theory. Central to our approach is the development of network-based representations that model social media interactions as complex networks. We explore various network representations, focusing particularly on the use of maximum entropy models, to identify statistically significant interactions and filter out noise within these networks. The research also integrates Natural Language Processing (NLP) techniques and temporal analysis to uncover more subtle forms of coordinated behavior and information diffusion.

We provide a broader context for the research by examining the historical evolution of information ma- nipulation, illustrating the shift from traditional media to digital platforms and the corresponding changes in manipulation tactics, and how these can be detected and countered. Next, we explore the concepts of graph randomization and maximum entropy models, leading to the creation of MaxEntropyGraphs.jl, a Julia package specifically designed for efficient graph randomization and analysis. This tool offers several benefits, includ- ing high performance and versatility, making it suitable for a wide range of applications beyond social media network analysis.

The effectiveness of the methodologies is demonstrated through the analysis of real-world disinformation campaigns, including an analysis of datasets from the Twitter information operations report, and a detailed study of a Belgian disinformation case. These studies not only validate our methodologies but also reveal in- sightful patterns and tactics used in modern disinformation campaigns. The thesis concludes by underscoring the increasing challenges posed by the evolving landscape of information manipulation, particularly with the rise of generative AI, and highlights the need for interdisciplinary collaboration and continuous adaptation of detection strategies.}},
  author       = {{De Clerck, Bart}},
  language     = {{eng}},
  pages        = {{XX, 185}},
  publisher    = {{Ghent University. Faculty of Economics and Business Administration}},
  school       = {{Ghent University}},
  title        = {{Identifying online misinformation and disinformation spreading using graph theory}},
  year         = {{2024}},
}