Identifying Extremism In Online Forums

June 10, 2020

One of the greatest developments in human history was the creation of the internet. Never before had so many people been connected, opportunities available, and ideas considered. However, the open format of the internet has also allowed for bad actors, such as right-wing extremists (RWEs), to congregate and exchange ideas with potentially deadly consequences.

Identifying digital signs of extremism is a growing priority for many law enforcement agencies at the local, state, and federal level. However, with over 4.5 billion active internet users – over half of the world’s population – manually reviewing all possible signs of extremism is impossible. Dr. Ryan Scrivens, an assistant professor in the School of Criminal Justice at Michigan State University, may have found a solution: AI and algorithms that incorporate criminal career measures.

A recent study published by Dr. Scrivens (pictured below) focused on a popular right-wing extremist forum Stormfront. In order to develop the algorithm that would serve as the study’s base, Dr. Scrivens looked for online discussions that related to RWEs’ top three adversary groups: members of the Jewish, Black, and LGBTQ+ communities. For each of these groups, a set of keywords – which included slur words and racial epithets – were used to identify content of interest. Traditional criminal career measures (i.e., frequency, seriousness, and duration) were then incorporated into the algorithm to identify RWE posting behaviors.

The results the algorithm yielded were three main types of RWE posting behaviors on Stormfront: High Intensity Posters, High Frequency Posters, and High Duration Posters.

High Intensity Posters were those who did not post a high volume of content but the messages were powerful and included calls to action against their adversaries.

High Frequency Posters were those who posted a high volume of content and who’s posts were more akin to rants about their adversaries rather than calls to action.

High Duration Posters were those who spent the most time posting about their adversaries and were the most dedicated to the “cause."

Dr. Scrivens’ findings revealed that High Intensity Posters appeared to be the opinion leaders in the online community, as their posts were powerful and persuasive, with much of their content advocating for violence against the Jewish community. High Frequency Posters, on the other hand, were not as persuasive to their peers and did not target any one adversary group in particular, but the amount of extremist content that they posted in the forum was noteworthy. Finally, High Duration Posters spent the most time posting in the forum and appeared to have the deepest seeded beliefs about the communities they hated.

Going forth, Dr. Scrivens hopes that this research will inform future risk factor frameworks used by law enforcement and intelligence agencies to identify credible threats online. He says that investigating radical posting behaviors within an already radical online community may provide practitioners with new insights into what constitutes online behaviors worthy of future attention. Dr. Scrivens says that this research is more important now than ever because “people engaging in violent extremism are connecting and communicating online. Identifying these individuals through their posting behaviors may help prevent future physical violence.”