Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A political radicalization framework based on Moral Foundations Theory

Ruben Interian

TL;DR

This paper addresses political radicalization in online communities by moving beyond speech-only analysis to a dual framework that combines Moral Foundations Theory with interaction-network structure. It defines a multidimensional relevance scale using four foundations (Fairness, Ingroup loyalty, Authority, Purity) as proxies for radicalization, while excluding Care, and operationalizes two measurement axes: speech-based foundation relevance and structure-based signals (Ingroup loyalty via $d_i$ on modularity and Authority via Partial Dominating Set). The authors apply a Pareto-frontier MCDA approach to identify non-dominated, highly radicalized groups, and validate the framework on four large Brazilian election Twitter datasets, finding a dominant right-leaning Pareto-optimal community in the post-election period, with observed mismatches between behavior and speech for Ingroup loyalty. The study demonstrates that network structure can reveal radicalization dynamics not always captured by content alone, suggesting practical utility for platform monitoring and risk assessment, while acknowledging methodological limitations like resolution in modularity-based community detection and context-dependent interpretation of Ingroup behaviors.

Abstract

Moral Foundations Theory proposes that individuals with conflicting political views base their behavior on different principles chosen from a small group of universal moral foundations. This study proposes using a set of widely accepted moral foundations (Fairness, Ingroup loyalty, Authority, and Purity) as proxies to determine the degree of radicalization of online communities. The fifth principle, Care, is generally surpassed by others, which are higher in the radicalized groups' moral hierarchy. Moreover, the presented data-driven methodological framework proposes an alternative way to measure whether a community complies with some moral principle or foundation: not evaluating its speech, but its behavior through interactions of its individuals, establishing a bridge between structural features of the interaction network and the intensity of communities' radicalization regarding the considered moral foundations. Two foundations may be assessed using the network's structural characteristics: Ingroup loyalty measured by group-level modularity, and Authority evaluated using group domination for detecting potential hierarchical substructures within the network. By analyzing the set of Pareto-optimal groups regarding a multidimensional moral relevance scale, the most radicalized communities are identified among those considered extreme in some of their attitudes or views. The application of the proposed framework is illustrated using real-world datasets. The radicalized communities' behavior exhibits increasing isolation, and its authorities and leaders show growing domination over their audience. There were also detected differences between users' behavior and speech, showing that individuals tend to share more 'extreme' ingroup content than that they publish: extreme views get more likes on social media.

A political radicalization framework based on Moral Foundations Theory

TL;DR

This paper addresses political radicalization in online communities by moving beyond speech-only analysis to a dual framework that combines Moral Foundations Theory with interaction-network structure. It defines a multidimensional relevance scale using four foundations (Fairness, Ingroup loyalty, Authority, Purity) as proxies for radicalization, while excluding Care, and operationalizes two measurement axes: speech-based foundation relevance and structure-based signals (Ingroup loyalty via on modularity and Authority via Partial Dominating Set). The authors apply a Pareto-frontier MCDA approach to identify non-dominated, highly radicalized groups, and validate the framework on four large Brazilian election Twitter datasets, finding a dominant right-leaning Pareto-optimal community in the post-election period, with observed mismatches between behavior and speech for Ingroup loyalty. The study demonstrates that network structure can reveal radicalization dynamics not always captured by content alone, suggesting practical utility for platform monitoring and risk assessment, while acknowledging methodological limitations like resolution in modularity-based community detection and context-dependent interpretation of Ingroup behaviors.

Abstract

Moral Foundations Theory proposes that individuals with conflicting political views base their behavior on different principles chosen from a small group of universal moral foundations. This study proposes using a set of widely accepted moral foundations (Fairness, Ingroup loyalty, Authority, and Purity) as proxies to determine the degree of radicalization of online communities. The fifth principle, Care, is generally surpassed by others, which are higher in the radicalized groups' moral hierarchy. Moreover, the presented data-driven methodological framework proposes an alternative way to measure whether a community complies with some moral principle or foundation: not evaluating its speech, but its behavior through interactions of its individuals, establishing a bridge between structural features of the interaction network and the intensity of communities' radicalization regarding the considered moral foundations. Two foundations may be assessed using the network's structural characteristics: Ingroup loyalty measured by group-level modularity, and Authority evaluated using group domination for detecting potential hierarchical substructures within the network. By analyzing the set of Pareto-optimal groups regarding a multidimensional moral relevance scale, the most radicalized communities are identified among those considered extreme in some of their attitudes or views. The application of the proposed framework is illustrated using real-world datasets. The radicalized communities' behavior exhibits increasing isolation, and its authorities and leaders show growing domination over their audience. There were also detected differences between users' behavior and speech, showing that individuals tend to share more 'extreme' ingroup content than that they publish: extreme views get more likes on social media.
Paper Structure (13 sections, 4 equations, 4 figures, 1 table)

This paper contains 13 sections, 4 equations, 4 figures, 1 table.

Table of Contents

  1. Introduction
  2. Methods
  3. Choosing primary moral foundations for measuring radicalization
  4. Relevance of moral foundations using group speech
  5. Interaction networks
  6. Group-binding loyalty and isolation
  7. Authority and hierarchy
  8. Integrating the relevance scales
  9. Results
  10. Datasets and communities
  11. Radicalization by network structural features
  12. Radicalization by group speech measuring
  13. Conclusions

Figures (4)

  • Figure 1: Group-level modularity. Example of a graph with three groups, with the black group being much more internally cohesive and isolated.
  • Figure 2: Example of a fully hierarchical structure of relations.
  • Figure 3: Radicalization analysis using network structure. The Pareto frontier is calculated for two structural features. The color indicates the group leaning (red for left-leaning and blue for right-leaning), and the shape indicates the Pareto efficiency of the solution (circle O for Pareto-efficient and X mark for non-Pareto-efficient).
  • Figure 4: Radicalization analysis using four group speech features. Communities are represented as connected line segments, and vertical lines represent moral foundations.