Abstract

Uncovering and understanding how influence networks push propaganda and disinformation within the wider news ecosystem remains a difficult challenge that requires tracking and characterizing how narratives spread across thousands of fringe and mainstream news websites. Using 18 months of daily news article scrapes from 1,003 unreliable news (e.g. twisted.news), 1,012 mixed-reliability websites (e.g., theepochtimes.com), and 2,061 reliable news websites (e.g., washingtonpost.com), the finetuned Matryoshka embedding, hierarchical Reciprocal Nearest Neighboring clustering, and zero-shot stance detection, we isolate and quantify the relationships between unreliable, mixed-reliability, and reliable news outlets. We show that by utilizing the stances of website articles toward particular entities and network inference-based tools, we can track slanted propaganda networks and identify the most influential websites in spreading particular attitudes, not only on fringe websites but within the broader media ecosystem, helping the reporting and fact-checking of propaganda and disinformation.

Bio: Hans is a rising 5th year Ph.D. student at Stanford University supervised by Professor Zakir Durumeric and researching in the Empirical Security Research Group. His research focuses on natural language processing, computer security, and the spread of misinformation online. His research is supported by the Meta/Facebook Ph.D. Research Fellowship and the National Science Foundation Graduate Research Fellowship. Hans completed two Masters’ degrees in Computer Science and in Statistics with the Daniel M. Sachs Scholarship at the University of Oxford. Hans completed his undergraduate degree in Electrical Engineering at Princeton University.

Zoom link: https://cam-ac-uk.zoom.us/j/83115049986?pwd=6W5bzFb49HcCbWqz6HR3tRhpVxubTb.1