A makeshift memorial outside Tree of Life synagogue in Pittsburgh on Oct. 29 after a deadly shooting. (Matt Rourke/AP)

Just a month ago, a gunman opened fire in Pittsburgh’s Tree of Life synagogue, killing 11 worshipers. The alleged shooter’s anti-Semitic ideology had simmered for years in online forums, where he found company and support among white supremacists and others in the far right. His last social media post before the shootings read, “I can’t sit by and watch my people get slaughtered. Screw your optics, I’m going in.” That was on Gab, a social media network that held itself out as a “free speech alternative” and was known for its users’ extremist views.

Experts quickly connected the alleged shooter, a truck driver named Robert Bowers, with Gab’s anti-Semitism by reviewing his posts, which were publicly available but have since been removed from the site, which itself was shuttered for several days after the attack but is once again available. When the Southern Poverty Law Center analyzed Bowers’s posts, it identified central themes that included “white genocide” and “globalism” — a buzzword often referring to anti-Jewish conspiracy theories. Similar analyses of his posts presented equivalent conclusions.

We have a unique set of data that includes not just all of Bowers’s Gab posts but also all Gab posts since the site launched in 2016. The daily use of Gab is climbing. When the Pittsburgh shooting occurred, we recognized that our data could tell us more about both the accused shooter and his network — and possibly offer clues about who might escalate from online hate to violence.

What did the alleged Pittsburgh shooter talk about on Gab, and with whom?

We used a custom API to scrape every post, reply and quote from every user on Gab, reaching back to the site’s launch in 2016. Our database includes all Gab posts, even those users subsequently deleted from the site. We used text-analysis methods in Python to analyze the massive volume of data.

Since joining Gab in January 2018, Bowers contributed 998 posts and replies to others’ posts. With text analytics of his nearly 1,000 posts, replies and quotes, we found overwhelmingly anti-Semitic ideations.

Bower’s two most commonly used words were “kike” and “jew,” often associated with terms like “vile” and “degenerate,” and many other anti-Semitic references as well. For example, he frequently used the hashtag “#everysingletime,” referring to an alt-right Jewish conspiracy theory in which Jews are behind orchestrated efforts to attack white culture and values.

Gab has “groups” in which members with similar interests hold conversations. Many of Bowers’s posts were in these groups, especially the “Guns of Gab,” “Supremacy,” “Immigration” and “Anti-white” groups, which reinforced his interest in firearms and white-supremacist philosophy.

Bowers boasted about 600 friends and followers on Gab but was far less active than many in his network. Judging by the number of messages sent among members, Bowers was tightly connected with a select few members of this network and loosely connected with many.

Comparing Bowers to the typical Gab user

We compared Bowers’s activity and his social network to the larger Gab community. Our analysis suggests that while Bowers does not perfectly represent the rest of Gab’s members, substantial similarities exist.

We gathered the 200 most commonly used hashtags on Gab to compare to Bowers’s. The hashtags included #altright and #whitegenocide, themes in Bowers’s posts. Bowers’s posts and the most popular Gab hashtags both included references to political conspiracy theories such as #qanon and #releasethememo.

But Bowers’s posts and replies differ in his nearly exclusive focus on anti-Semitic ideas. In the rare times he touched on other topics, those were general alt-right political discussions about white supremacy and related themes. The average Gab user was much more focused on general current political topics, with hashtags including #maga, #trump, #fakenews and #politics.


Most common Gab hashtags, August 2016 through October 2018

Robert Bowers’s most commonly used words on Gab

The larger network active on Gab

Gab conversations are heavily laden with alt-right themes. As you can see in the illustration above, since 2016, one of the hashtags most commonly used on Gab has been #altright.

Other common Gab hashtags from the same data set suggest broader anti-immigration themes, including #buildthewall, #bansharia and #banislam. These hashtags were among the most common in a set of more than 30 million posts — indicating that these sentiments are pervasive. As you can see, Gab also overflows with political conspiracy theory discussions, with such common hashtags as #pizzagate, #draintheswamp and #deepstate. Gun rights activism shows up in hashtags such as #2a and #nra.

Users have posted roughly 30 million comments on Gab since its 2016 launch. Below you can see that the volume of Gab activity by posts, replies and quotes has grown over time.


Gab usage, 2016 through 2018

White supremacy and anti-Semitism have endured in the United States. The Anti-Defamation League documented a 57 percent increase in anti-Semitic incidents from 2016 to 2017. FBI hate crime statistics for 2016 show that more than half of faith-based crimes were against Jews.

After the August 2017 Unite the Right rally in Charlottesville, many expected the alt-right to retreat from physical gatherings and return to digital realms. Instead, alt-right gangs such as the Proud Boys, the Atomwaffen Division and their followers, have become bolder, engaging in street violence at rallies, bomb plots and killings.

Our analyses show that Bowers is not alone in his hatred. Gab has roughly 450,000 users, and while many do not espouse anti-Semitism, a sizable number do and regularly connect online to share their ideology. Our analyses also show that these individuals are not difficult to find or monitor.

The challenge comes in knowing when they will escalate from online speech to real-world action.

Matthew Phillips (@mdphill1_uncc) is an assistant professor of criminal justice and criminology at the University of North Carolina at Charlotte. 

Arunkumar Bagavathi is a final-year PhD student in the computer science department at the University of North Carolina at Charlotte. His research interests include data mining, network analysis and cloud computing.

Shannon E. Reid (@mourningshannon) is an assistant professor of criminal justice and criminology at the University of North Carolina at Charlotte.

Matthew Valasik (@MattValasik) is an assistant professor of sociology and criminology at Louisiana State University.

Siddharth Krishnan is an assistant professor of computer science at the University of North Carolina at Charlotte. His research uses Web mining, computational social science and applied machine learning to build explanatory and predictive models of actions of large groups of people and societies.

The authors thank Ryan Wesslen and Gabriel Fair, PhD students in computer science at University of North Carolina at Charlotte, for their assistance in collecting the data.