JMIR Hum Factors. 2026 Apr 8;13:e78236. doi: 10.2196/78236.
ABSTRACT
BACKGROUND: Social media is one of the most accessible and extensive sources of data for tracking and understanding public reactions to COVID-19 policies. Cultural differences between the United States and Japan have resulted in highly distinctive policies and public reactions in each country.
OBJECTIVE: This study aims to analyze the public opinions surrounding COVID-19 mask mandate through 1,102,876 and 560,873 geo-tagged tweets from the United States and Japan during the period from 2020 to 2022. We conducted 3 stages of analysis-relevance to COVID-19 masks, stance for or against masking, and whether the tweets indicate users wearing masks-to understand individuals’ stance towards the mask mandate and their actual mask-wearing behavior.
METHODS: We adopted a semisupervised approach to enhance BERT (Bidirectional Encoder Representations from Transformers) classification results due to data imbalance, which were then visualized through time series and map representations.
RESULTS: In the United States, our data showed that individuals with a bachelor’s degree or higher, as well as those living in states with higher household incomes, are positively correlated with positive attitudes toward mask-wearing. In contrast, in Japan, those with higher education levels or individuals aged 65 years and older were positively correlated with tweets categorized as having a stance against the mask mandate. Key events in Japan, such as the announcement of the state of emergency and the Olympics, served as major triggers for the number boost in public opinion.
CONCLUSIONS: Our analysis of over 1.6 million tweets from the United States and Japan revealed that public opinion shifted notably in response to major events and policy changes during the COVID-19 pandemic. While some trends align with previous research, correlations with education, age, and income suggest that social media data may reflect underlying societal divisions and algorithm-driven biases.
PMID:41950353 | DOI:10.2196/78236
