CSH PostDoc Hannah Metzler will give this online talk within the framework of the Data Science/ Computational Social Science Seminar Series of the University of Michigan.
This seminar takes place on April 28 from 12 to 1 pm Eastern Daylight Time.
Registration for the DS/CSS events at umsi.info/DSCSS.
In this talk, I will first present two case studies on the relationship between social media measures for emotions at the population-level with emotions reported in surveys. They included both dictionary- as well as machine-learning based emotion measures, and one included data from two different social media platforms. Overall, we found that social media emotion measures closely tracked emotions reported in surveys from the UK and Austria in a time period before and during COVID-19. The results show that daily and weekly social media indicators can represent emotional trends in societies at large, but also highlight the need for further validation studies.
Second, I will present a large-scale study on collective emotions during the early COVID-19 outbreak. Social media emotion measures based on tweets in six different languages showed strong and enduring increases of anxiety and sadness expressions, together with decreases in anger expressions in 18 countries around the world. These changes were in part related to increases in COVID-19 cases and the stringency of measures against the spread of the virus. Taken together, these studies illustrate that social media emotion measures can provide added value in addition to representative surveys, in particular during unexpected crisis events.