COMPUTATIONAL SOCIAL SCIENCE SEMINAR – Quantifying the Social Debates of Anti-Vaccination on Twitter – Yuan

October 6, 2017 @ 3:00 pm – 4:30 pm
Center for Social Complexity Suite 3rd Floor, Research Hall, Fairfax Campus
Karen Underwood


Xiaoyi Yuan, PhD Student
Computational Social Science Program
Department of Computational and Data Sciences
George Mason University

Quantifying the Social Debates of Anti-Vaccination on Twitter

Friday, October 6, 3:00 p.m.
Center for Social Complexity Suite
3rd Floor, Research Hall


Measles is one of the leading causes of death among young children. In many developed countries with high measles, mumps, and rubella (MMR) vaccine coverage, measles outbreaks still happen each year. Social media has been one of the dominant information sources to gain vaccination knowledge and thus has also been the focus of the “anti-vaccine movement”. This talk is about two of my recent research projects on this topic. The first one will be introduced briefly, which is an agent-based model demonstrating how a small amount of online anti-vaccine sentiment could have the power of increasing the probability of measles outbreaks significantly. This research inspired me to investigate details of communicative pattern of “anti-vacciners” by analyzing a large twitter dataset (660892 tweets) after the California Disneyland measles outbreak in 2015. This second research has two main parts: first, in order to identify “anti-vacciners”, I used supervised learning to label each tweet as either positive, neutral, or negative opinion towards vaccination. The linear support vector machine model shows good performance on this dataset with an accuracy score of 72% on test data. Second, Louvain’s method for community detection of the retweet network shows the common pattern of social media communities; i.e., overall fragmented but with a few large communities. By investigating the opinion distribution in big communities, however, I discovered that they are highly overlapped, especially within “anti-vacciners”, meaning that they have more frequent communication within their own opinion group than with others. What’s useful for health communication strategies is to look further into the brokers–those who stand between two or more communities. At the end of the talk, I will address details of analyzing the brokerage as well.