Analyzing information sharing behaviors during stance formation on COVID-19 vaccination among Japanese Twitter users

PLoS One. 2024 Dec 31;19(12):e0299935. doi: 10.1371/journal.pone.0299935. eCollection 2024.

Abstract

To prevent widespread epidemics such as influenza or measles, it is crucial to reach a broad acceptance of vaccinations while addressing vaccine hesitancy and refusal. To gain a deeper understanding of Japan's sharp increase in COVID-19 vaccination coverage, we performed an analysis on the posts of Twitter users to investigate the formation of users' stances toward COVID-19 vaccines and information-sharing actions through the formation. We constructed a dataset of all Japanese posts mentioning vaccines for five months since the beginning of the vaccination campaign in Japan and carried out a stance detection task for all the users who wrote the posts by training an original deep neural network. Investigating the users' stance formations using this large dataset, it became clear that some neutral users became pro-vaccine, while almost no neutral users became anti-vaccine in Japan. Our examination of their information-sharing activities during a period prior to and subsequent to their stance formation clarified that users with certain types and specific types of websites were referred to. We hope that our results contribute to the increase in coverage of 2nd and further doses and following vaccinations in the future.

MeSH terms

  • COVID-19 Vaccines* / administration & dosage
  • COVID-19* / epidemiology
  • COVID-19* / prevention & control
  • East Asian People
  • Humans
  • Information Dissemination*
  • Japan
  • Social Media*
  • Vaccination Hesitancy / psychology
  • Vaccination Hesitancy / statistics & numerical data
  • Vaccination* / psychology

Substances

  • COVID-19 Vaccines

Grants and funding

This research was conducted as part of "COVID-19 AI & Simulation Project" run by Mitsubishi Research Institute commissioned by Cabinet Secretariat, JAPAN. The methods for analysis were developed with support from JST CREST Grant Number JPMJCR19A4 and JSPS KAKENHI Grant Number JP21H03445. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.