Justifying Dyadic Harm in Political Manifestos

This study examines whether linguistic indicators of harm and integrative complexity differentiate political manifestos associated with violence. Drawing on the theory of dyadic morality, which conceptualizes moral judgment as structured around relationships between intentional agents and vulnerable patients, the present research analyzes how moral representations of harm relate to violent outcomes in real-world texts. A corpus of 267 political documents — including violent, ideologically extreme, and moderate manifestos — was analyzed using dictionary-based linguistic measures and automated integrative complexity scoring. Marker prevalence was calculated per 1,000 words, and regression models were used to assess differences across categories and predictors of violence classification. Results indicated that dyadic harm language was the most consistent predictor of violence. Violent manifestos exhibited higher levels of dyadic harm framing than moderate texts, and dyadic harm significantly predicted whether a manifesto was associated with violence. In contrast, integrative complexity did not predict violence in either baseline or full models. Justification for violence and martyrdom narratives also did not significantly predict violence, and justification language was more prevalent in moderate texts than in violent or extreme manifestos. Hopelessness was highest in ideologically extreme texts and was associated with a lower likelihood of violence. These findings suggest that the transition from ideological extremity to violence is more strongly associated with how moral situations are represented — specifically in terms of agent–patient harm — than with how beliefs are justified or cognitively structured. By extending harm-based models of moral cognition to political manifestos and clarifying the limits of integrative complexity as a predictor, this study contributes to a more precise understanding of the relationship between moral representation and violent behavior.