Dresden 2026 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

SOE: Fachverband Physik sozio-ökonomischer Systeme

SOE 11: Polarization

SOE 11.1: Talk

Wednesday, March 11, 2026, 15:45–16:00, GÖR/0226

Accurate mean-field predictions for cognitively grounded social influence dynamics with confirmation bias — •Sven Banisch — Karlsruhe Institute of Technology

Collective opinion formation in human groups is often modeled through cognitively rich agent-based dynamics, yet the resulting high-dimensional systems typically resist analytical treatment. Here we show that a class of cognitively grounded argument-exchange models with confirmation bias admits a low-dimensional and fully tractable mean-field representation. The key idea is to project pairwise agent rules onto an influence response function and then derive a two-compartment mean-field system whose Jacobian separates into a mean mode and a polarization mode. This yields a simple stability test that diagnoses the onset of symmetry breaking. We recover the consensus-to-polarization tipping point and identify a second threshold at which polarized equilibria become fully stable. Across all parameter regimes, the mean-field bifurcation structure accurately reproduces the behaviour of the full agent-based model. Our results provide a general mechanism-level framework for understanding how cognitive biases shape macroscopic opinion patterns, and demonstrate that high-dimensional social influence dynamics can exhibit simple and universal phase-space structure amenable to analytical classification.

Keywords: opinion dynamics; model reduction; bifurcation analysis; polarization; confirmation bias