Introduction
The COVID-19 pandemic was accompanied by an "infodemic"—a massive surge of information that complicated public health responses. This research investigates the psychological coupling between official government communications and public sentiment on Twitter in India. By analyzing lead-lag relationships, the study seeks to understand how official narratives influence collective mood and where public concerns precede government action.
Methodology and Data
The researchers introduced the CoronaIndiaDataset (CID), comprising 5.6 million tweets and official bulletins from six Indian states. To analyze this data, they utilized a customized version of Empath, a deep-learning-based psychometric tool.
Standard sentiment analysis often misinterprets medical terms; for example, "testing positive" is linguistically positive but contextually negative. The team manually refined categories like medical emergency, health, and government to ensure accuracy.
To determine causality, the study employed Granger’s Causality tests, identifying whether one time series (bulletins) could predict the future values of another (Twitter sentiment) better than the series' own past.
Key Insights: Who Leads Whom?
The analysis revealed distinct patterns in how information and emotion flow between the state and the public:
- Government as the Medical Lead: Official bulletins significantly "led" the conversation regarding Medical Emergency, Death, and War. When the government released data or directives, the public response followed.
- Public as the Hygiene Lead: Conversely, the public led the government on topics of Hygiene. Discussions about handwashing and sanitization peaked on Twitter before becoming a primary focus in official state bulletins.
- The Impact of Lockdown: Public optimism actually rose in mid-April as the national lockdown began to slow the virus's doubling rate, suggesting that decisive government action can positively influence public morale.
- Official Stoicism: While public sentiment fluctuated with fear and confusion, government bulletins maintained a strictly clinical and administrative tone, almost entirely lacking emotional markers.
Regional Variations
Discourse was not uniform across the country. The study identified "state-specific personalities" in the psychometric data. For example, tweets from Delhi and Punjab were heavily focused on government actions, whereas West Bengal and Tamil Nadu showed a higher concentration of medical emergency discussions.
Conclusion
The study highlights a complex feedback loop between official communication and public sentiment. While the government remains the primary driver of medical information, the public often identifies practical concerns like hygiene earlier. To facilitate real-time monitoring, the authors launched COVibes, an interactive dashboard for visualizing these psychometric insights.
A major limitation remains the focus on English-language data in a multilingual nation. Future research aims to incorporate regional languages to provide a more comprehensive view of India's psychological landscape during crises.
Summary updated