Feeling Angry? Blame Your Friend.
Have you ever felt happy or upset after hearing that your friend or close family member was feeling this way? Turns out, you’re not alone.
Researchers in China recently conducted research to understand how members of a social network emotionally influence one another. In their paper, “Anger Is More Influential Than Joy: Sentiment Correlation in Weibo,” the authors discuss how they constructed an interaction network from tweets crawled from Weibo (a very popular Chinese social media site similar to Twitter) between April 2010 to September 2010. By building an interaction network, the researchers ensured that an edge would only connect two nodes (users) if the users replied, retweeted or mentioned each other in their tweets at least 30 times. This ensured that the users included in the data set had strong ties, as opposed to weak ones. Thus, the process of following a user is simply not enough to warrant the establishment of an edge. The resulting data set was composed of 70 million tweets, from 200,00 users, from which a weighted constructed graph was created. This analysis is relevant to Networks, Crowds, and Markets’ discussion about strong ties, weak ties and social graphs.
The authors looked at how four major emotions- anger, joy, sadness and disgust were present in these tweets, not through the text itself, but through the emoticon, which is used ubiquitously by users. 3.5 million tweets with emoticons from the data set were used as a training corpus. A Bayesian classifier was created which then used as training data to determine the sentiment of the remainder of the data, which did not necessarily have emoticons. Upon identifying the sentiment of users’ specific tweets, the researchers were able to analyze how emotions proliferate through a network.
The results of the study were compelling- while sadness and disgust did not correlate between users who were connected to one another, joy did experience some correlation. But hands down, anger was the emotion that had the strongest correlation. The authors found that anger’s influence could be felt within three hops in a neighborhood, meaning that anger as an emotion can be fairly contagious. Given that social media has been considered a tool for the freedom of expression and political agency, perhaps this research provides some insights into why Twitter, Facebook and Weibo are effective in helping to rally groups of individuals with a cause or those who want to engage in some form of activism. We know that anger can propagate across a social network regardless of whether technologies are used. But technologies can significantly speed up this process and expand the influence of anger so that it radiates outward. As discussed in Network, Crowds and Markets, homophily is the idea that we are similar to our friends. These finding suggest that emotion homophily doesn’t exist between random users, but that emotion is closely connected to social ties.
One thing that is left for us to question is the extent to which weak ties are emotionally influenced by one another. If a pair of users who have weak ties are surrounded by a neighborhood who are strongly correlated for anger, how likely will it be that they will also exhibit anger in their tweets? What might the role of bridges and local bridges be in disseminating certain emotions to other nearby components?
Future research could explore how other emotions move through social networks- one particularly interesting context in which to examine this would be to see how fear spreads during crisis events such as natural disasters or large scale civil unrest. Applications such as Ushahidi, could then be designed to better account for the emotional patterns which directly affect how users in a network behave.