I explore how misinformation spreads through websearch, with a focus on how and why unreliable websites can rank highly in search engines. I additionally introduce models that can help search engines improve the reliability of search results. I first explore the limitations of Google's current efforts to combat low-reliability and low-relevance content in its search engine. I then demonstrate both how owners of unreliable websites can manipulate structural authority patterns present in the webgraphs, and introduce content-agnostic Graph Neural Networks to detect unreliable websites using these signals. I then explore the content-based approaches that unreliable websites use to rank on search engines and again propose models for detecting them. I also explore how the presence of unreliable websites on SERPs impacts Google's AI overviews. I then combine the content and webgraph with social media mentions of unreliable websites to create a realistic model of the paths that users can take to reach an unreliable website in hopes of better understanding the signals present in each. Finally, I combine the takeaways from the previous chapters into a conceptual framework that will be integrated into a simulated information environment exercise, where misinformation experts will use the work developed in this thesis to characterize and identify misinformation and narratives.

202 pages

Thesis Committee:
Kathleen M. Carley (Chair)
Uttara Ananthakrishnan
Jamie Callen
Francesca Tripodi (The University of North Carolina at Chapel Hill)

Nicolas Christin, Head, Software and Societal Systems Department
Martial Hebert, Dean, School of Computer Science

Return to: SCS Technical Report Collection
School of Computer Science

This page maintained by reports@cs.cmu.edu