Using Social Media Data to Capture Emotions Before and During COVID-19

During the COVID-19 pandemic, our social lives moved online to a larger extent than ever before, as opportunities for face-to-face social contact in daily life became increasingly limited. Because of this, the authors of the World Happiness Report have turned their focus on what can be learned about people’s emotional experiences and well-being from analyzing text data on social media. 

Such data is relevant for emotion research because emotions are not only internal experiences but are often social in nature. Given their valuable social function, emotions are regularly shared with other people, influencing other people’s emotions. For instance, happiness may spread through social networks and give rise to clusters of happy and unhappy people.

Because social media continuously captures communication between millions of people over long periods, researchers have been able to collect this data and trace the emotions and well-being of individuals and societies at new scales and resolutions. 

Three case studies presented in the Report prove that emotion measures based on social media postings can track emotions at a society-wide level. These aggregate measures seem to be more accurate for measuring affective experiences at shorter time scales, with correlations highest for short-lived emotions reported daily and lowest for more slowly changing measures of well-being like satisfaction with life. 

This collected social media data can support various research questions for which survey data are unavailable, such as retrospective analyses, crisis research, or studies on populations that are hard to reach with surveys. The authors have presented one example for crisis research, using indicators of emotional well-being in 18 countries during the COVID-19 outbreak. During the first five weeks of the COVID-19 outbreak, they’ve observed strong initial increases in expressions of anxiety on Twitter, associated with the growth in cases and the stringency of measures. A bit later, social media measures of emotional expressions indicated a gradual increase in sadness and decrease in anger, which began when stringency measures included strict lockdowns. 

Anxiety gradually decreased once measures had been implemented, suggesting that people habituated to the new circumstances or felt reassured by their governments’ actions. Anger expressions dropped as discourse on social media shifted away from politically polarized discussions and focused on COVID-19. Sadness seemed more strongly associated with the effects of social distancing measures on people’s personal lives and only linked to deaths by COVID-19 as these became more prevalent.

The correlation studies presented in the Report suggest that social media data reveal information about the emotional well-being of residents of these countries during this early pandemic stage. Taken together, social media emotion data provide added value in addition to representative surveys.

The correlations researchers observed in the UK study were in the range of correlations between surveys, suggesting that social media data are suitable as a complementary source of information on emotions. Social media and survey data may potentially contribute some unique information to predict outcomes like suicide hotline calls, hospital visits, police calls, or overdose rates. Future research could explore if combining these two data sources could help better predict and respond to such important outcomes.

Keep Reading about the World Happiness Report 2022 findings. Exploring the Biological Basis for Happiness