Embedding-aware Polarization Management in Signed Networks

TL;DR

EPM introduces an embedding-based polarization measure grounded in effective resistance and a structure-aware mitigation strategy via localized augmentation through structurally balanced intermediary nodes and demonstrates that EPM effectively mitigates polarization while preserving task-relevant network structure.

Abstract

Signed network embeddings (SNE) are widely used to represent networks with positive and negative relations, but their repeated use in downstream analysis pipelines can inadvertently reinforce structural polarization. Existing polarization measures are largely designed for unsigned networks or rely on predefined opinion states, limiting their applicability to embedding-based analysis in signed settings. We propose EPM, a unified polarization management framework that jointly measures and mitigates polarization in the embedding space. EPM introduces an embedding-based polarization measure grounded in effective resistance and a structure-aware mitigation strategy via localized augmentation through structurally balanced intermediary nodes. Experiments on real-world signed networks demonstrate that EPM effectively mitigates polarization while preserving task-relevant network structure. The codebase of EPM is available at https://github.com/JeonghanSon/EPM-Embedding-aware-Polarization-Management.

Figures (4)

• Figure 1: Temporal evolution of structural polarization in signed networks. Across all datasets, structural polarization consistently increases as networks grow, suggesting that polarization can accumulate as signed networks evolve.
• Figure 2: Overview of the proposed EPM framework. The framework consists of (a) an embedding-aware polarization measurement component that quantifies polarization from signed network embeddings, and (b) a gray-zone-based mitigation component, which reduces polarization via indirect structural augmentation while preserving the original network structure.
• Figure 3: Effect of the negative edge scale factor $\boldsymbol{\eta}$ on polarization. Lower $\boldsymbol{\eta}$ strengthens negative edges in the ER geometry.
• Figure 4: Sensitivity to pair-selection hyperparameters (EQ4). Lower $\boldsymbol{\tau}$ and larger $\boldsymbol{d_{\max}}$ yield stronger polarization reduction with predictable trade-offs in Macro-F1.