Predicting Polarization Using Personality Traits - Calibrating an Agent-Based Model on Survey Data

Resumen

We examine whether classic opinion-dynamics models can be strengthened by grounding their core parameters in psychological domain knowledge. First, we fielded a preregistered survey experiment in Germany (N = 1,804) that mimics a single ABM updating step. Participants reported justice-relevant traits, empathy, ambiguity tolerance, justice sensitivity, and moral indignation, then engaged with counter-attitudinal content. From these data we estimated generalized linear models linking traits to two Hegselmann–Krause parameters: the confidence bound c (openness to differing opinions) and the self-influence weight m (inverse to susceptibility to influence). Empirically, a subset of theorized links was supported: cognitive empathy widened c, moral indignation narrowed c, and victim-oriented justice sensitivity increased m. We then calibrated a bounded-confidence ABM with these mappings and evaluated it against four canonical polarization findings. The empirically informed model substantially improved qualitative replication of group extremization and signed directional change under mixed evidence, matched the backfire contrast as good as the baseline, and, like the baseline, failed to reproduce a superlinear polarization slope. These results show how justice-focused traits, particularly empathy and moral indignation, provide a parsimonious, data-driven bridge from micro-level psychology to macro-level polarization dynamics, while also revealing limits that motivate extensions (e.g., identity cues and framing) in future work.

Áreas y palabras clave

Cómo citar (APA 7)

Geppert, R., & Maes, J. (2025). Predicting Polarization Using Personality Traits - Calibrating an Agent-Based Model on Survey Data. https://doi.org/10.31234/osf.io/42a7c_v1