The Future of Violent Extremism Research – 5 Recommendations Based on a Machine Learning Analysis of 34,000+ Articles

To discern trends in violent extremism research, we present a machine-learning analysis of over 34,000 articles published since the early 1900s. We identify two primary waves of research and, different to previous reviews, a clear diversification in studied groups, contexts, and topics. Less than 20% of articles employ methodologies that are conducive to drawing causal inferences. While more studies are using experimental and longitudinal methods, this increase is outpaced by the rise of methods that preclude the assessment of causality. Nuancing previous reviews, at the broader field level, violent extremism research is profoundly multidisciplinary, with political science, international relations, psychology, history, and law emerging as the “Big Five” contributors. At the study level, contributions from single disciplines remain the norm. While single-author contributions have rapidly declined in favor of team-based research, one-time contributing authors remain consistently high over time. To enhance future violent extremism research, we make five recommendations: (1) prioritizing methodologies that allow for causal inferences; (2) incorporating state-based violent extremism perspectives also in individual-level research; (3) increasing the utilization of big data by interdisciplinary teams; (4) increasing the focus on developmental research to understand early life influences; and (5) initiating more interdisciplinary work at the study level.