Unifying the Extremes: Developing a Unified Model for Detecting and Predicting Extremist Traits and Radicalization

Area of Science:

• Computational Social Science
• Psychology
• Sociology

Background:

• Online ideological movements increasingly form extremist factions.
• Existing research on radicalization is often ideology-specific, lacking generalizable characterization.
• Quantifying extremism across diverse online communities remains a challenge.

Purpose of the Study:

• To propose a novel, generalizable method for extracting and analyzing extremist discourse.
• To develop a framework for quantifying extremism at user and community levels.
• To identify a generalized psychosocial model of extremism.

Main Methods:

• Extraction and analysis of extremist discourse from online ideological communities.
• Focus on verbal behavioral signatures of extremist traits.
• Development of a framework quantifying extremism based on 11 distinct factors ('The Extremist Eleven').

Main Results:

• Successfully characterized ideologically diverse online communities across the 11 extremist traits.
• Predicted user entry into the incel community up to 10 months prior with high accuracy (AUC > 0.6, increasing to ~0.9).
• Observed that users maintain extremist traits upon joining forums, distinct from general online discourse.

Conclusions:

• Introduced a holistic, cross-ideological approach to studying extremism.
• The 'Extremist Eleven' framework offers a generalized psychosocial model for quantifying online extremism.
• The method demonstrates predictive power for identifying individuals at risk of radicalization.