Click here for free access to our latest coronavirus/COVID-19 research, commentary, and news.

Support nonprofit science journalism

Science’s extensive COVID-19 coverage is free to all readers. To support our nonprofit science journalism, please make a tax-deductible gift today.

According to an analysis of Twitter data, social unrest like the riots and marches in Baltimore can be predicted ahead of time.

According to an analysis of Twitter data, social unrest like the riots and marches in Baltimore can be predicted ahead of time.

AP Photo/Matt Rourke

Can unrest be predicted?

The broken glass and burned wreckage are still being cleared in the wake of the riots that convulsed Baltimore's streets on 27 April. The final trigger of the unrest was the funeral of a 25-year-old African-American man who had died in police custody, but observers point to many other root causes, from income inequality to racial discrimination. But for a few researchers who are studying Baltimore's unrest, the question is not the ultimate causes of the riot but its mechanism: How do such riots self-organize and spread? One of those researchers, Dan Braha, a social scientist at the New England Complex Systems Institute in Cambridge, Massachusetts, has been collecting data from Twitter that spans the riot from buildup to aftermath, part of a larger study of social media and social unrest around the world.

Q: What can you learn about the Baltimore riots from social media?

A: The protesters are mostly teens who use social media routinely. The riots that started around 3:30 p.m.—ignited by messages on social media urging high school students to “purge”—spread within 3 hours around the city. It's interesting to see the pattern of spread, much like forest fires, spreading in clusters and locally. The riots, in my view, could easily spread also across other cities in the United States where racial tensions are high and are close to a tipping point.

Q: How do you spot this tipping point?

A: There are three basic data from tweets: location, time, and intensity. First, you create a network of communication, where the nodes are people or groups. Then, you correlate the communication patterns with characteristics such as gender, political attitudes, and sentiment towards the political issues. For example, in the Arab Spring communication network, you saw a lot of communication going on among Arab-speaking people related to protesting, going out on the street, meeting at a certain time. It is the intensity of the communication—how many messages over time—that predicts what is to come. There is a tipping point, a threshold beyond which the fire starts. The threshold depends on the particular people, culture, and region. It is shaped by economic and political conditions. But the age of the people is a critical factor. Young people are the drivers of the events.

Q: How can you tell that social media enables the spread of unrest, rather than simply serving to comment on it?

A: The rioting happens in just hours and sometimes minutes of the communications on social media, with thousands of people pouring out into the streets. Without social media, it wouldn't happen.

Q: How does Baltimore's rioting compare to that in Ferguson, London, and elsewhere?

A: In terms of the communications and patterns of spread, they are remarkably similar. Even if the causes are different, there seems to be a universal pattern to civil unrest.

Q: Is it possible to predict social unrest?

A: It is absolutely possible. By tracking social media, you know exactly where and when to send the riot police. You can even inject misinformation into this system [to head off riots]. You want to inhibit the damage of unrest, but in a democratic society, we must ask ourselves if we want to do this. We have to find the right balance.