More is More: Scaling up Online Extremism and Terrorism Research with Computer Vision

Scholars and practitioners investigating extremist and violent political actors’ online communications face increasingly large information environments containing ever-growing amounts of data to find, collect, organise, and analyse. In this context, this article encourages terrorism and extremism analysts to use computational visual methods, mirroring for images what is now routinely done for text. Specifically, we chart how computer vision methods can be successfully applied to strengthen the study of extremist and violent political actors’ online ecosystems. Deploying two such methods – unsupervised deep clustering and supervised object identification – on an illustrative case (an original corpus containing thousands of images collected from incel platforms) allows us to explain the logic of these tools, to identify their specific advantages (and limitations), and to subsequently propose a research workflow associating computational methods with the other visual analysis approaches traditionally leveraged.