Emotion in Social Media
Galen Thomas Panger. Emotion in Social Media. Ph.D. dissertation. Advisor: Steven Weber. University of California, Berkeley. 2017.
The role of emotion in social media has been the subject of considerable research and media attention. But while stereotypes about the emotional profile of status updates — that they are overly-positive, or overly-angry — have solidified, evidence remains circumstantial and indirect. Further, although researchers have made numerous efforts to use the emotions we express in status updates to make inferences about our emotional lives — generating national happiness indices, predicting mental illnesses and evaluating emotional outcomes of experimental interventions — little is known about the validity of these inferences at the individual level, and researchers have largely ignored the impact of self-presentation and privacy concerns on validity. Finally, while debate continues about the emotional impacts of browsing social media in the course of day-to-day life, researchers have focused only on a limited set of emotions, rather than investigating the range of human emotion.
To address these issues, I present three analyses regarding (1) the emotions we express in social media, (2) what can be inferred about our emotional lives in general based on how we express ourselves in social media, and (3) the emotional experience of browsing social media. I conduct experience sampling for one week with participants in a Facebook sample (N = 344) and Twitter sample (N = 352), gathering data about their day-to-day emotional lives. I then compare this data to participants’ ratings of the emotional contents of their most recent status updates so as to reveal the distinct emotional profile of status updates and address questions regarding the validity of inferences. Data from experience sampling is also used to reveal the emotional experience of browsing social media.
In the first analysis, across a broad spectrum of emotions, I find status updates to be largely similar in emotional profile to emotional life in general, though Facebook posts are more positive on average, and tweets are more negative. Both Facebook posts and tweets exhibit higher levels of emotions associated with activation (energy, alertness) and lower levels of emotions associated with deactivation (drowsiness, sleepiness) than emotional life in general. In the second analysis, I find that the emotions we express in status updates have a low-moderate correlation with day-to-day emotional life, suggesting that efforts to infer emotional life from the emotions we express in status updates have some validity. The association is weaker, however, for individuals higher in attention to self-presentation and privacy in the Facebook sample, and disappears in both the Facebook and Twitter samples when a popular sentiment analysis program known as Linguistic Inquiry and Word Count (LIWC) is used to measure the emotional contents of status updates.
Finally, in the third analysis, I find that the emotional experience of browsing social media is characterized primarily by deactivation (by winding down), with a slight tilt toward negative emotion. While browsing Facebook, on average, is associated with slightly elevated feelings of envy, browsing Twitter is associated with relief of envy. Further, results suggest little potency for theories of emotional contagion in social media. Overrepresented emotions in Facebook posts and tweets do not tend to be reflected in the emotional experience of browsing Facebook or Twitter. Among many implications, the results of this dissertation suggest that social media is not whipping people into a frenzy on average, but rather, is predominantly calming. While counterintuitive, this result is robust and is found with both Facebook and Twitter.