Analyzing social media extremism

“Extremism” is a word we sometimes throw around casually, using it to label any actions or beliefs that seem, well, extreme. Other popular phrases include “extremist ideology,” and “violent extremism.”

When you hear the word “extremists” or “extremism,” what do you imagine? Probably something along the lines of ISIS or white nationalists, groups that are explicitly extremist. But these based on examples are insufficient because it becomes incredibly difficult to categorize extremism when we move away from the ends of the spectrum.

How do we objectively define this nebulous term beyond “I know it when I see it?”

That’s the problem that J.M. Berger, a postgraduate research student at Swansea University’s School of Law and research fellow at VOX-Pol, investigates, working a lot with homegrown terrorism and social media extremism to try and clarify this social phenomenon into a more objective framework. In a talk given this past Tuesday, Berger delved into his ideas on what distinguishes extremism in the 21st century.

Drawing upon historical analogues — such as Pope Innocent the Third, Hernan Cortez, and Edmund Ruffin — he created a definition for extremism based on social identity theory, which draws on in-group and out-group dynamics, where the in-group is the extremist group and the out-group is comprised of everyone else. He also defines the term “ideology” as a set of documents detailing how members of the in-group should interact with or treat members of the out-group. Thus, an extremist ideology is simply a set of rules on how extremists in an extremist in-group should treat those in the out-group. As Berger notes, because extremist groups are very concerned with their legitimacy, they take great caution in clearly defining the boundaries for their members.

With the foundation of social identity theory, extremism can be broken down into three key parts: identity, crisis, and solution. Each of these three components must exist in some form for extremist groups to form, and they progress sequentially.

The identity defines who is in the in-group and who is in the out-group, and this division follows from the very strict guidelines set forth by those in the in-group. There is also a level of nuance where the out-group is divided into the eligible and ineligible out-group, where the in-group appeals to the eligible out-group for recruits and supports. In the case of ISIS, the eligible out-group would include Muslim males, as there is a potential for recruitment among that population.

Once the in-group and out-group are established, the in-group defines a kind of crisis that forces confrontation of the out-group. For example, a popular conspiracy theory for anti-Semites is the idea that Jewish people secretly control the world. These crisis narratives can range in severity from conspiracy — the idea that the out-group secretly controls outcomes for the in-group — all the way to apocalypse, where the out-group is seen as a threat to all of humanity.

The final piece is the solution, where extremist groups present a way to resolve the crisis. Solutions can range from non-violent harassment all the way to genocide, with the severity and extremeness increasing along the spectrum.

So why is it so important to reduce extremism to objective components? It opens up the possibility of combating extremism online with artificial intelligences rather than human monitoring. A team of humans could never sort through all the hundreds of thousands of text posts and messages sent online every day. By training AIs to recognize the elements of extremism, we can monitor extremist dialogue before it becomes an issue. As Berger demonstrated later in his presentation, all of these abstract ideas can be reduced to concept maps and frameworks, which in turn can be fed to natural language processing systems and threat detection AIs who can do the heavy lifting.

After the talk, the audience raised some intriguing questions, such as if reactions to extremism can themselves be considered extreme. Another great question was whether or not the current administration has played a role in escalating extremism. Without naming any names, Berger explained that the current political rhetoric fulfills the identity and crisis portion of the extremism cycle — such as Trump categorizing immigrants as the out-group threat — without providing any solutions. When people are left to their own solutions, the results can be catastrophic.

This thought-provoking lecture explained how vague concepts can be reduced to functional definitions, so long as the definitions are carefully thought out and supported by legitimate evidence. As we move into an age where subjectivity is used frequently as a defense for hate, this approach is vital to the viability of our society. Only by identifying the harmful agents can we curtail their impact, and machines can help us achieve what manpower alone can not.