Project Description #
This research investigated the anti-vaccine movement on Twitter (now X) through computational social network analysis, aiming to understand who spreads anti-vaccine misinformation, what communities form around this rhetoric, and what persuasive strategies are employed. The study collected 19,250 tweets based on 21 hard-coded anti-vaccine hashtags identified through domain knowledge and news articles, all gathered on a single day (January 24, 2022) due to Twitter API rate limitations. From this dataset, 5,362 unique public accounts were extracted, and a follower network was obtained from a random sample of 293 public users. The research employed multiple computational methods: follower-count and hashtag-frequency mapping to identify opinion leaders, Infomap community detection on a three-layered hierarchical network to characterize anti-vaccine communities, and Latent Dirichlet Allocation (LDA) topic modeling to uncover dominant rhetorical strategies. Key findings revealed that the most influential anti-vaccine users appeal to common relatable narratives, such as family, religion, financial interests, and political allegiance. Community detection identified geographically distributed communities across the US, Canada, and the UK that were homogeneous in account age and follower size but connected by weak ties that nonetheless facilitated effective information spread. Rhetorical strategies centered on government control concerns, conspiracy theories about official narratives, medical misinformation (particularly regarding myocarditis), economic anxieties, calls for political action, and concerns about children’s health. The project was awarded First Place at the 2023 CIC Student Paper Challenge.
My Role #
As a co-author alongside Sydney McCormack and Ziran Gu, I contributed to the full research pipeline under the guidance of Professor Bernard Koch. My work involved designing and executing the data collection methodology using the R programming language, including hashtag-based tweet filtering and extraction of unique user IDs from the resulting dataset. I participated in building follower network visualizations to identify the top 10 opinion leaders by mapping follower count against anti-vaccine hashtag usage, and analyzed the common narratives (family, religion, money, and political allegiance) employed by these influential accounts. I contributed to the application of Infomap community detection to model information flow across the follower network and helped implement single-topic LDA (via Python) to surface dominant rhetorical strategies across the tweet corpus. I co-authored the resulting paper, which won first place at the 2023 CIC Student Paper Challenge, and presented the research findings at the associated conference.
Publication #
The paper itself will be uploaded to this site shortly. I presented on its findings at the virtual 2024 NEBDHub Student Research Symposium webinar, viewable at this YouTube link or using the embed below.