Quantifying COVID-19 Content in the Online Health Opinion War Using Machine Learning

Abstract

Misinformation about the COVID-19 virus is spreading rapidly online, and it might be very harmful. In this work, we apply machine learning to measure COVID-19 discourse among online anti-vaccination ("anti-vax") activists. The anti-vaccine movement, we have found, is COVID-19 has received less attention from the anti-vaccination camp ("anti-vax") than vaccination itself. However, the anti-vax community displays a wider variety of "_favors" of COVID-19 topics, and so can appeal to a larger proportion of people looking for COVID-19 guidance online, including those who are leery of a mandatory fast-tracked COVID-19 vaccine and those who are interested in alternative treatments. Therefore, it seems that the anti-vaccine movement will be more successful in attracting new supporters in the future than the pro-vaccine movement will. This is worrying since it means that the globe will fall short of giving herd immunity against COVID-19, leaving nations vulnerable to future resurgences of the disease. We provide a mechanistic model that provides insight into these findings and may be useful for gauging the potential efficacy of intervention efforts. Our method is salable, so it may be used to the pressing issue of social media platforms needing to sift through vast amounts of online health misinformation and deception.