Tweeting terror: Understanding extremism on social media
Extremist organizations like the Islamic State group (ISIS or ISIL) and Al-Qaeda have recruited more than 25,000 foreign fighters from over 100 countries, according to the United Nations. Their numbers increased 70 percent from 2014 to 2015. How are these groups able to attract so many followers so quickly?
Although extremist groups appear to advocate for a return to the Dark Ages, they show a sophisticated mastery of modern social media technology. When a terrorist organization disseminates a message through social media, it doesn’t just reach the citizens of one country or region.
“It has a global footprint,” says Hasan Davulcu, an associate professor in Arizona State University’s School of Computing, Informatics, and Decision Systems Engineering
When the Islamic State group posts an execution video, for example, every “share” accelerates its spread across the Internet. Once an online message has gone viral, it becomes extremely difficult to counter.
At least 46,000 Twitter accounts are used by Islamic State supporters, according to a Brookings Institution report (March 2015).
The social side of security
Davulcu has been addressing this challenge with a global, interdisciplinary team of experts. In 2009, ASU received a highly competitive Minerva Project grant from the U.S. Department of Defense. The Minerva Project initiative seeks to understand the social, cultural, behavioral and political forces that shape various regions of the world.
Davulcu and Mark Woodward, an associate professor in ASU’s School of Historical, Philosophical and Religious Studies, led the five-year project. They collaborated with researchers from around the world who specialize in communications, anthropology, religious studies, political science and sociology. Their task was to create a tool that could track and analyze online messages produced by radical and counter-radical Muslim groups in Southeast Asia, West Africa and Western Europe.
Building this tool was a daunting prospect. It would require incredibly sophisticated software that had the ability to analyze multiple languages and multiple cultural contexts. Davulcu is skilled at mining data online from websites and social media platforms, but he needs to know what to look for. That guidance comes from social scientists providing extensive on-the-ground cultural knowledge of each of the nine countries included in the project: the UK, Germany, France, Niger, Nigeria, Senegal, Malaysia, Singapore and Indonesia.
For example, Woodward, an anthropologist, has traveled to Indonesia, Malaysia and Singapore since the late 1970s. He builds rapport and gains local knowledge by talking with a wide range of people in everyday settings, such as schools or mosques. This allows him to learn about the issues that are important to a community and gauge their experiences in dealing with radical extremist groups.
“These are the people that have to deal with the likes of Al Qaeda every day,” Woodward says.
Over the course of the Minerva Project grant, researchers identified all the known radical and counter-radical Muslim organizations in the nine countries of interest. Social scientists provided cultural knowledge about the groups, while Davulcu scoured the web for Twitter feeds, Facebook posts, recorded speeches and other related content. Together, the research team began to construct an extensive map of the Muslim religious extremist landscape in different regions of the world.
In addition to many scientific publications, presentations and awards, a product of this ambitious effort was a tool called LookingGlass. LookingGlass is a computer program that maps and analyzes the radical and counter-radical Muslim groups based on variables such as religion, political affiliation, levels of tolerance and engagement in violent behavior. For the first time, researchers could visualize and track hot spots of social networks, their narratives and activities, and the sociocultural, economic and political drivers behind them.
“The DOD was impressed over and over again by the fact that we can map multiple contexts, multiple languages, and through a single methodology,” Davulcu says.
Predicting viral posts
The project was such a success that in June 2015, ASU was awarded a second Minerva Project grant, led by Davulcu and Paulo Shakarian, an assistant professor in the School of Computing, Informatics, and Decision Systems Engineering. Once again, the researchers are collaborating with scientists from different disciplines and different countries. Their goal is to discover why certain messages on social media spread more rapidly than others and develop ways to predict when something will go viral.
That capability could be a game-changer. It would allow governments to effectively counteract the online recruiting effort of radical groups. The challenge is that researchers don’t have many examples of how a viral post unfolds within a given data set, such as Muslim extremists. In addition, messages online go viral for very different reasons.
In his initial research, Shakarian has discovered some social factors that are associated with information cascades, or viral messages. For example, he says that the number of distinct communities someone follows, rather than the number of individual people, is more predictive of how that person will be influenced online.
To understand this concept, imagine that you have a large social network made up entirely of former college classmates. They all start posting on social media about a company they believe is engaging in corrupt practices. They share their intention to boycott the company and all of its products. After seeing these messages, you might be persuaded to do the same.
Now, imagine instead that your social network consists of a few people from college, a few from high school, some former coworkers, some current ones and some people you know through a shared hobby. All of these people from different groups start posting about the corrupt company and their plans to stop buying its products. These groups live in different parts of the country and have never met, yet they’ve come to the same conclusion about the company. How would this influence your opinion?
Shakarian’s research suggests that your collective network in the second scenario would be more persuasive than the first.
“Because then you start thinking, ‘Oh, well I’m getting this from all kinds of different places,’” he says.
This is a strategy that extremist groups have used to increase their reach. They recruit many people all over the world to flood social media with their messages. This effort is hard to counter and attempts to do so could backfire.
“You’re showing that it’s important enough to be responded to,” Shakarian explains.
However, if the messages aren’t refuted, they run the risk of going viral. Having the ability to predict their impact is critical.
Through his initial research on the spread of messages across people and communities, Shakarian has been able to predict whether a post will go viral with about 50-60 percent accuracy. This is significant, because guessing is only correct about two percent of the time. To further increase his predictive accuracy, Shakarian will need more information about the people and communities that are producing these messages. That’s where the social scientists come in.
“Qualitative people with their very deep insights can carry the computer types and the quantitative people from the first ‘aha’ moment to the next ‘aha’ moment,” Davulcu says.
Social scientists are on the front lines, which is important in places like the Middle East, where national boundaries are rapidly changing. The Minerva Project grants awarded to ASU have been successful because the research team has taken an interdisciplinary approach to this global problem.
To describe it in another way, Davulcu invokes an old Indian parable: “You are not going to understand an elephant like this by looking at parts of it.”