EXTREMISMThe Promise—and Pitfalls—of Researching Extremism Online

By Heather J. Williams, Alexandra T. Evans, and Luke J. Matthews

Social scientists are fascinated with social media. These new technologies have altered how humans interact with one another, providing new windows to observe how humans communicate, learn, and build relationships. The result has been a “data gold rush (PDF),” with scholars mining for insights into human behavior.

For the subset of us studying extremism, social media has a particular allure. Previously, doing such research demanded physically infiltrating groups to observe how they operated. Now, from laptops or cell phones, researchers can monitor extremists as they network, recruit, radicalize, communicate, and mobilize.

But while online spaces are key enablers for extremist movements, social media research hasn’t provided many answers to fundamental questions. How big of a problem is extremism, in the United States or around the world? Is it getting worse? Are social media platforms responsible, or did the internet simply reveal existing trends? Why do some people become violent?

We have few answers because this research is easy to do poorly and hard to do well. The challenges fall into three buckets: users, platforms, and content.

Users Don’t Equal People
Like the ever-controversial Twitter owner Elon Musk, extremism researchers want to know how many real people exist behind the voluminous number of user accounts. Researchers must account for the possibility that many of these are inauthentic. In addition to automated bots, a single user can control many accounts, share an online identity with others, or participate in a “cyborg” account comprised of both bots and real people.

Anonymity also intertwines with the user issue. Some social media platforms attempt to verify users’ identities, but many require only an email address, and some don’t require any registration at all. This makes it difficult to determine demographic details—like gender, age, or location—that might produce meaningful conclusions about who is involved in extremist movements. And extremists likely are drawn to platforms that provide anonymity.

Researchers have developed some tools to distinguish inauthentic accounts and infer basic demographic information. Still, these methods are imperfect and those involved in extremist networks have many incentives to hide or obscure their identities. At the same time, not everyone chooses to be on social media, so the available data may be skewed by an overrepresentation of certain demographic groups.

These constraints complicate answering some of the most basic questions: How big a problem are we facing? When responding to extremism, who do we need to help?