Structural Analysis of the COVID-19 Infodemic: Motif-Based Detection of Echo Chambers and Geopolitical Hijacking in Global News Networks Using GDELT

Abstract

The COVID-19 pandemic produced a global infodemic that evolved structurally across time. This paper presents an automated informatics pipeline that transforms high-velocity GDELT 2.0 Global Knowledge Graph (GKG) data into a mathematically validated model of narrative evolution across five pandemic milestones. The pipeline integrates elite-domain authority filtering across 21 globally recognised news sources, BERTopic-based semantic node extraction, and Heterogeneous Information Network (HIN) construction. Two network motifs, Narrative Stars (broadcast hubs) and Sociosemantic Triads (echo chambers), are enumerated and validated against 1,000 degree-preserving null models using Monte Carlo permutation testing. BERTopic consistently outperforms the LDA baseline (Cv = 0.58) with coherence scores above Cv = 0.70 at all milestones, peaking at Cv = 0.7777 during M2 Lockdown. Motif analysis reveals a statistically significant Star-to-Triad crossover, with Stars peaking at Z = 156.65 in M2 and Triads peaking at Z = 210.55 in M5 (p < 0.001). Longitudinal Louvain-Jaccard tracking identifies near-zero community survival (J = 0.0075 at the M2-to-M3 transition), confirming structural collapse rather than gradual evolution. During M4 Delta, betweenness centrality shifts toward geopolitical entities, providing quantitative network evidence of narrative displacement. The paper contributes a reproducible topological framework for infodemic surveillance.