Twitter communities are associated with changing user's opinion towards COVID-19 vaccine in Japan

Despite extensive research studying the opinions on vaccination since the outbreak of COVID-19, the dynamics of temporal opinion shifts of individuals and their possible origins have been rarely studied. Here, we explore the possible influence of social interactions (retweet network) on individual's opinion shift related to vaccinations based on large-scale Twitter data in Japan. We use an opinion score which calculates the fraction of pro-, neutral and anti-vaccine tweets to measure the dynamic changes of individual opinions, and identify statistically significant communities based on retweet network. By tracking individual's dynamic opinion and its community affiliation, our study highly suggests that the opinion shifts are largely influenced by the user's Twitter community. That is, if users are within the anti-vaccine (or pro-vaccine) community, they exhibit a significantly higher likelihood of changing their position and adopt an anti-vaccine (or pro-vaccine) stance. We also find that the anti-vaccine community's influence appears to persist longer than the pro-vaccine community.