Collective Emotions during the COVID-19 Outbreak

Abstract

The COVID-19 pandemic has exposed the world's population to sudden challenges that elicited strong emotional reactions. Although investigations of responses to tragic one-off events exist, studies on the evolution of collective emotions during a pandemic are missing. We analyzed the digital traces of emotional expressions in tweets during five weeks after the start of outbreaks in 18 countries and six different languages. We observed an early strong upsurge of anxiety-related terms in all countries, which was stronger in countries with stronger increases in cases. Sadness terms rose and anger terms decreased around two weeks later, as social distancing measures were implemented. Positive emotions remained relatively stable. All emotions changed together with an increase in the stringency of measures during certain weeks of the outbreak. Our results show some of the most enduring changes in emotional expression observed in long periods of social media data. Words that frequently occurred in tweets suggest a shift in topics of conversation across all emotions, from political ones in 2019, to pandemic related issues during the outbreak, including everyday life changes, other people, and health. This kind of time-sensitive analyses of large-scale samples of emotional expression have the potential to inform mental health support and risk communication.

Social media emotion macroscopes reflect emotional experiences in society at large

“…These results are robust when considering autocorrelation and heteroskedasticity for all cases except for the last case of dictionary-based positive emotions. The case of dictionary-based anxiety also shows some signs of non-stationarity due to the large spike in the time series at the onset of the pandemic, as already reported with similar methods 27 , but results are still comparable when using robust estimators that consider outliers and in permutation tests (see SI for more details). Signals measured with the supervised method are comparable to the dictionary-based method for sadness and anxiety, but for the case of the emotion happy in YouGov, the supervised method gives significantly stronger positive correlations than the dictionary-based method in both the historical and prediction periods.…”

“…Following our pre-registration (https://aspredicted.org/blind.php?x=r89nv2), we analyzed the text of tweets using the anxiety, sadness, and positive affect dictionaries of LIWC through Brandwatch slightly adapted as in our previous work 27 . This way, we obtain a daily count of tweets within the sample that contain at least one emotional term, which we then average over weekly windows aligned with the YouGov survey.…”

Collective emotions during the COVID-19 outbreak.

“…Data analysis was done in R Version 3.6.1 (R Core Team, 2018), using the packages from the Tidyverse (Wickham et al, 2019), as well as lme4, scales, cowplot, ggrepel, and gridExtra (Auguie, 2017; Bates et al, 2015; Slowikowski, 2019; Wickham & Seidel, 2019; Wilke, 2019). All R code and data are available at https://osf.io/736kc/ (Metzler et al, 2021). Only the machine learning analysis for the robustness check was done in Python Version 3.9.2, using packages Transformers (4.5.1) and Torch (1.8.1+cu102).…”

‘I Enjoy Having Someone to Rant to, I Feel Like Someone is Listening to Me’: Exploring Emotion in the Use of Qualitative, Longitudinal Diary-Based Methods

“…Thus, using metaphors as an analytic tool, Stanley et al (2021) have demonstrated that emotional experiences of COVID-19 converged around several deeply held negative emotions: (a) grief, (b) disgust, (c) anger and (d) fear. Meanwhile, Metzler et al (2021) analysed tweets from 18 countries during the first 5 weeks of the outbreak and observed a strong early upsurge in anxiety-related terms. Further, sadness terms rose and anger terms decreased around 2 weeks later as social distancing measures were implemented.…”