Skip to main content

Verified by Psychology Today


When Do Peaceful Protests Turn Ugly?

Researchers find that moral attitudes, as measured in tweets, predict violence.

Peaceful protest is a cornerstone of democracy. Demonstrations offer a way for constituents to communicate concerns, focus attention on issues, and promote change. Unfortunately, protests can quickly turn violent. Although many people find protesting for causes that they care about liberating, it’s safe to say that most people—including police officers and government officials—would prefer to avoid violent protests.

Using AI and Twitter, researchers at the Brain and Creativity Institute at USC discovered that people are more likely to promote violence when they are moralizing the issue about which they are demonstrating, according to a new paper in Nature Human Behavior. Furthermore, this effect is moderated by moral convergence, or the extent to which a person thinks others share similar moral attitudes.

“Extreme movements can emerge through social networks,” co-author Morteza Dehghani told USC News. “We have seen several examples in recent years, such as the protests in Baltimore and Charlottesville, where people’s perceptions are influenced by the activity in their social networks. People identify others who share their beliefs and interpret this as consensus. In these studies, we show that this can have potentially dangerous consequences.”

In this study, researchers analyzed 18 million tweets posted during the Baltimore protests of 2015, which focused on the death of Freddie Grey, a victim of police brutality. Over several weeks, these protests were punctuated by periods of peace and periods of violence, enabling the researchers to assess the association between social media rhetoric and violent incidents.

To label these tweets as either “moral” or “not moral,” the team developed a deep neural network using 4,800 tweets coded for moral content.

“We focus on morality because once a protest is sufficiently moralized, it becomes an issue of right and wrong instead of mere personal preference,” the researchers wrote. Dehghani and colleagues compared the moral content of the labeled tweets with arrest rates, which served as an imperfect proxy for violence.

Senyuk Mykola/123RF
Source: Senyuk Mykola/123RF

The researchers also ran three behavioral experiments to elucidate the relationship between moralization and perceived moral convergence. The participants in these analyses were primed with accounts from the violent Unite the Right rally in Charlottesville, Virginia, in 2017.

“Although the behavioral experiments measure the acceptability of using violence at protests instead of protest behavior,” wrote the researchers, “the fine-grained text analysis of the Baltimore protests and the three behavioral experiments taken together aim to provide converging evidence for our hypotheses using both real-life protests and self-reported attitude measures.”

The team first found that moral rhetoric nearly doubled on violent protest days. On further analysis, as hypothesized, they found that hour-level tweets predicted violence. In other words, as the number of moral tweets went up, so did subsequent arrest counts.

“A rise in violence at protests may thus reflect the increasing moralization and polarization of political issues in online echo chambers,” the researchers wrote.

In their series of behavioral analyses, the researchers found that participants were more likely to endorse violent protest when they moralize an issue. The extent to which they promote violence, however, is based on whether others share their outlook.

Although violence at protests is well publicized, very little research has been done on this topic. The current study has important implications.

First, this research may help decision makers better predict how to allocate resources to prevent a protest from turning violent.

Second, the researchers suggest that “decreasing the moralization of attitudes and diluting the perception that others agree with one’s moral position may attenuate the rise of the acceptability of violence.” In other words, if we can decrease the extent to which people morally converge and select themselves into networks of like-minded individuals, then we may reduce the acceptance of violence during protests.

Finally, although moral outrage and moral convergence are required for violent protest, the researchers point out that other factors—such as violent proclivities among protestors and the nature of the issue being protested—also play a role.


Mooijman M et al. “Moralization in social networks and the emergence of violence during protests.” Nature Human Behaviour. May 23, 2018.

More from Naveed Saleh M.D., M.S.
More from Psychology Today