Using Machine Learning to Compare Provaccine and Antivaccine Discourse Among the Public on Social Media: Algorithm...

Area of Science:

• Public Health
• Social Media Analysis
• Communication Studies

Background:

• Antivaccine content on social media persistently increases, often linked to vaccine hesitancy.
• Provaccine campaigns struggle to engage and persuade the public effectively.
• Prior research suggests antivaccine content's topic diversity drives higher public engagement.

Purpose of the Study:

• To compare discursive topics in pro- and antivaccine social media content.
• To explain engagement disparities between provaccine and antivaccine content.
• To identify how vaccine topics are framed using Entman's four framing dimensions.

Main Methods:

• Analysis of 39,962 tweets from a public social media site.
• Development of supervised classification for tweet categorization (provaccine, antivaccine, neutral).
• Application of unsupervised clustering for topic identification and qualitative content analysis for framing.

Main Results:

• Antivaccine topics exhibit greater distinctiveness than provaccine topics (t=2.30, P=.02).
• Antivaccine advocates utilize all four persuasive message frames.
• Provaccine advocates often neglect to include a clear problem statement in their messaging.

Conclusions:

• Distinctiveness of antivaccine topics contributes to higher engagement.
• Comprehensiveness of message frames influences vaccine debate impact on uptake.
• Urgent need to develop clear problem statements for provaccine content to mitigate antivaccine influence.