The radicalisation loop: A layered linguistic model of disengagement and re-engagement in an incel forum

Abstract Online engagement with extremist communities presents significant mental health and real-world risks. This study investigated how patterns of engagement within an incel forum relate to dynamic affective and cognitive linguistic markers. Using longitudinal linear mixed models and Linguistic Inquiry and Word Count (LIWC) on a dataset of 2,009,875 comments, spanning six years, we compared users who permanently left (Desisters), those who persisted (Persisters), and those who left but later returned (Returners). Whilst overall differences between groups were minimal, distinct psychological trajectories emerged: Desisters showed an increase in grievances and hostility over time, whereas Persisters showed more stable emotions and increased social clout. Returners showed lower desperation and frustration following a period of disengagement, yet their re-engagement was marked by a significant decline in analytic thinking and a persistence of hate, signalling re-radicalisation. These findings highlight a novel, layered model of online radicalisation, offering insights for developing targeted clinical interventions.