What makes online content viral?

Consider two news articles. The first article is the most-read Guardian story of 2015, about the Paris attacks (Philips & Rawlinson, 2015). The second article is the third most-read Guardian story of 2015, about flowing water on Mars (Sample, 2015). Whereas the two articles have attracted an amount of readers of similar order of magnitude, the first one has been shared almost 50,000 times, whereas the second time has been shared only 6 times. How is this possible? This effect can be explained by looking at emotional impact, at least that is what Berger and Milkman (2012) hypothesize in their paper ‘What makes online content viral’? The authors investigate the characteristics of news articles and their impact on the amount of email shares. In particular, they focus on the valence of an article and the emotions it evokes. This blog post discusses the research and points out its practical implications.

Two of the most-read items on theguardian.com have striking differences in the number of social shares

The authors hypothesize that positive content will get shared more than negative content. People often share content for the purpose of self-presentation and identity communication, and negative content may reflect negatively on the sender. Apart from valence, social transmission may also get shaped by differences in arousal, in a sense that content that evokes high arousal or activation will get shared more.

In order to investigate the hypothesized effects, two types of studies were combined. First, data from the New York Times’ news articles and their list of most emailed items were analyzed, controlling for practical utility, interestingness and surprisingness of the article content. Second, lab experiments were conducted to verify that observed differences in virality were the result of specific emotion and arousal.

The results confirm the authors’ expectations. In particular, articles evoking the high-arousal emotions awe (positive), anger or anxiety (negative) are more likely to go viral than articles evoking relaxation or sadness. Moreover, although not of core focus to this research, the control variables practical utility, interestingness and surprise were all found to be significant predictors of virality.
Returning to the comparison presented in the introduction, this suggests that the striking difference in social transmission was caused by the fact that the high-arousal emotions of anger and anxiety the Paris attack article evoked, outweighed the effect of the positivity of the article about Mars.

The paper’s findings have interesting implications for practice. Firstly, it shows how practitioners crafting marketing campaigns can make these more effective, in terms of social transmission and virality. They do not necessarily need to craft something positive, but can choose from aiming to evoke awe, anger or anxiety. In fact, positive but low-arousal evoking content is often even less viral than negative, high-arousal content. Moreover, incorporating practical utility or making sure the marketing campaign is interesting or surprising will result in more social transmission.
Secondly, the research’ results may help practitioners to estimate at an early stage the impact of negative publicity. All other things being equal, content evoking sadness is less likely to get widespread than content evoking anger or anxiety. Through these insights, companies can better determine the need to respond.

Content evoking high-arousal emotions gets shared more