News Release

US Islamist extremist co-offenders form close-knit groups driven by mutual contacts, homophily effects

Mutual contacts and the “birds of a feather” effect appear to play key roles in forming these close-knit clusters of extremists

Peer-Reviewed Publication


Extremists of a feather flock together? Community structures, transitivity, and patterns of homophily in the US Islamist co-offending network


US Islamist co-offending network. Note: isolated nodes are excluded. The network includes only individuals who had at least one co-offending relationship.

view more 

Credit: Schwarzenbach, Jensen, 2024, PLOS ONE, CC-BY 4.0 (

The formation of relationships within violent US Islamist extremist groups is highly driven by mutual contacts and the tendency for people to bond with others similar to themselves, according to new research. Anina Schwarzenbach, formally of Harvard University and the University of Maryland (currently affiliated with the University of Bern) and Michael Jensen of the University of Maryland present these findings in the open-access journal PLOS ONE on June 5, 2024.

Prior research on social structures within extremist networks have primarily explored outcomes, such as insurgency or conflict, with scant focus on how relationships first arise. The few studies on relationship formation have suggested that, unlike groups that are mostly profit-driven, terrorist groups prioritize trust-based relationships that increase security over those boosting efficiency.

Building on that prior work, Schwarzenbach and Jensen analyzed how trust-based co-offending relationships—relationships between extremists who commit ideologically motivated crimes either together or separately after receiving each other’s influence—form within networks of Islamic extremists radicalized within the United States. They analyzed data from the publicly accessible Social Networks of American Radicals (SoNAR) database, which includes court-record data on relationships between US Islamist offenders.

The researchers first applied algorithms for detecting community structures to SoNAR data, revealing that the US Islamist co-offending network consists mostly of small, separate, close-knit community clusters, with only a few larger communities.

Next, they applied a technique called exponential random graph modeling to SoNAR data to examine the potential roles of homophily—a tendency to bond with similar people—and transitivity—connection through mutual contacts—in the formation of these co-offender relationships. They found that relationships were shaped by mutual contacts, ideological affiliation, spatial proximity, and shared socio-cultural traits, suggesting that both homophily and transitivity help to drive co-offending relationships among US Islamist extremists.

The authors note several limitations to this work, including that their findings may not be generalizable outside of the US or to other kinds of extremist offenders. Still, they say, the findings point to several avenues for disrupting terrorist networks and preventing violence, such the importance of engagement-prevention programs and policing techniques that are informed by the local—as opposed to online— context in which trust-based co-offending relationships typically form.

The authors add: “The results underscore the significance of local connections and personal interaction in the mobilization of extremist activities. They suggest that combating terrorism requires a multifaceted and localized approach, combining efforts in the digital area with traditional police work at a local level.”


In your coverage please use this URL to provide access to the freely available article in PLOS ONE:

Citation: Schwarzenbach A, Jensen M (2024) Extremists of a feather flock together? Community structures, transitivity, and patterns of homophily in the US Islamist co-offending network. PLoS ONE 19(6): e0298273.

Author Countries: Switzerland, USA

Funding: This research was supported with a grant (grant number PP400PS 186706) from the Swiss National Science Foundation ( awarded to AS. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Disclaimer: AAAS and EurekAlert! are not responsible for the accuracy of news releases posted to EurekAlert! by contributing institutions or for the use of any information through the EurekAlert system.