Who Considers Terrorism Justifiable? ML Analysis Across 65 States (onlinelibrary.wiley.com)

Researchers used machine learning to analyze survey responses from 65 countries to identify who considers terrorism justifiable and which factors predict that view. By training predictive models on large cross‑national survey data, the team showed that only a minority endorse violence, but that endorsement is systematically associated with measurable demographic, socioeconomic and attitudinal features. Models highlighted predictors such as perceived political marginalization or injustice, low institutional trust, certain forms of religiosity and ideological alignment, and poorer socioeconomic status—while the relative importance of these predictors varied substantially between countries and regions. Technically, the study combined supervised learning (tree‑based classifiers and ensemble methods) with interpretable ML tools (feature importance and attribution analyses) and robust cross‑validation to test generalizability across contexts. The results matter to AI/ML and policy audiences because they demonstrate how predictive analytics can surface actionable risk factors for radicalization at scale, inform targeted prevention strategies, and reveal limits of one‑size‑fits‑all interventions. The paper also raises ethical and technical caveats: cross‑country heterogeneity, potential biases in survey sampling, privacy risks from predictive profiling, and the need for careful causal interpretation before using models in policy or surveillance.