In grey, the interactions of a social network, and in color, that network mapped through emotions (color coding is same as the figure below). Via Rui Fan et. al
Ever notice that whenever some clown writes a truly abhorrent, asinine blog post, it ends up being the most-shared article in your Twitter or Facebook feed that day? It's annoying as a reader—and as a journalist with at least a modicum of professional interest in having stories shared—to have everyone and their mother jump out of the digital woodwork to shout "Boo this man! [link]" when there's so many superior stories worth reading every day.
Hate-sharing seems like an undeniable force on the web. Now, a team of researchers have appeared to quantify it. According to work published on the arXiv pre-print server, anger is viral.
The researchers from Beihang University in Beijing looked at Weibo, China's Twitter-like social network that now boasts half a billion users. Weibo, like Twitter, is extremely important for both breaking news, politics, and citizen journalism, especially in a country that still has a heavy hand with the press. It's also a social network like any other, which means it's filled with plenty of jokes, joy, and social fluff.
The team, led by Rui Fan, set out to test whether or not homophily—human similarity, or as the team writes, the idea that "birds of a feather ﬂock together”—drives social connections in a network. As they explain, "beyond typical demographic features such as age, race, hometown, common friends and interest, homophily also includes psychological states, like loneliness and happiness." So, in other words, do Weibo users band together by their typical emotional state?
For their dataset, the team first gathered 70 million tweets from 200,000 users, all of whom had at least 30 interactions (e.g. had some sort of influence) during the collection period. They then decided to categorize them based on four emotions: anger, sadness, joy and disgust. How does one assign emotional states to that many tweets? Naturally, the team focused on emoticons.
They organized 95 common emoticons into those four groups, and categorized the 3.5 million tweets in their dataset using that scale. They then used that data—what words correspond with what emoticons, essentially—to create a bayesian classifier for the rest of the tweets. The team admits that it's a "simple but fast" system, but the "averaged precision of this classiﬁer is convincing."
Finally, the big moment: Which emotion would be most likely to create a reaction in kind, and then reactions from users farther and farther down the chain?
The above graph shows the correlation between the emotions of a tweet and its response as a pair of users move farther away in the social network. (An h of one are highly-connected users, h of 6 are very distant.) According to this data, anger moves most quickly through a network. As the authors write,
This suggests that anger could spread quickly and broadly across the network because of its strong inﬂuence to the neighborhoods in the scope of about three hops. Although the previous studies show that happiness is assortative in online social networks, but Figure 3 further demonstrates that the correlation of anger is much stronger than that of happiness. It means the information carrying angry message might propagate very fast in the network and this phenomenon is contrary to our intuition.
Joy is still a viral emotion, while sadness and disgust are much less so. But anger wins out. As Technology Review points out, a lot of this anger was found to be in relation to politics, both international and domestic. That's not surprising, as political problems tend to be popular as well as incite anger and frustration.
As Brian Fung at the Washington Post rightly notes, the phenomenon is only studied on Weibo, and thus we should be cautious in carrying over the same conclusions to other social networks in other countries. Anecdotally, it still makes sense. Anger, especially in relation to important issues, is highly shareable; we see it every day. Joy—and all those puppy pictures out there—are also highly shareable, but perhaps don't inspire the same passion as other topics.
I'd also venture to guess that sadness and disgust simply don't translate as well. People want to share things they're either passionate about or that they feel smart about sharing, and sad or disgusting things are harder emotions to fit into either of those categories. I guess the retweet barrier for "This guy is a shithead" is much lower than for "Life sux," perhaps because sadness and disgust don't jive with our carefully-manicured online images. But righteous indignation? The internet loves that.