Misleading the Covid-19 vaccination discourse on Twitter: An exploratory study of infodemic around the pandemic

Misleading vaccine content on Twitter had distinct linguistic patterns: The study identified specific features—including sentiment, hashtags, nouns, and pronouns—that differentiated misleading tweets from accurate ones, suggesting that false vaccine information uses recognizable language markers that could help identify it automatically.

Machine learning models achieved high accuracy in detecting vaccine misinformation: Multiple machine learning models successfully classified tweets as misleading or non-misleading with up to 90% accuracy, demonstrating that computational approaches can effectively flag false vaccine information at scale across large social media datasets.

Explainability is crucial for trustworthy misinformation detection: The researchers used SHAP (an explainable AI tool) to show why their models classified tweets as misleading rather than just providing predictions, highlighting that transparent, interpretable systems are important for responsible deployment of misinformation detection tools.

The study provides a foundation for actionable misinformation mitigation: By making their dataset and code publicly available on GitHub, the researchers enabled the broader research community to reproduce, validate, and build upon their findings—supporting collaborative efforts to combat vaccine misinformation rather than siloed approaches.

Misinformation about COVID-19 vaccines was widespread during a critical period: The analysis of 200,000 tweets over seven months (September 2020-March 2021) captured a pivotal moment in vaccine rollout, revealing the scale and nature of false information circulating when public trust and accurate information were most needed.