Hierarchical propagation networks¶
Analysis of information diffusion at multiple levels of granularity within social networks. Rather than examining only macro-level cascades (e.g., the global retweet tree) or micro-level interactions (e.g., replies to individual tweets), hierarchical approaches jointly model both levels to capture how claims propagate from their origin through amplification waves and then through localized discussion and debate.
Key concept¶
Information spread on social media exhibits structure at multiple scales: - Macro-level: Global propagation from news source → original tweets → retweets → cascading retweets (how far and wide a claim reaches) - Micro-level: Local conversations on individual posts or reposts (how deep discussions go, what sentiment and stances users express)
Hierarchical analysis treats these as complementary signals: macro-level features reveal how many users engage; micro-level features reveal what kind of engagement occurs.
Key papers in this wiki¶
- Shu et al. (2019) — Hierarchical Propagation Networks for Fake News Detection: Constructs hierarchical propagation networks from FakeNewsNet data; extracts structural, temporal, and linguistic features from both macro-level retweet cascades and micro-level reply trees; shows micro-level and macro-level features capture complementary information (F1 0.843/0.862 with both vs. 0.802/0.854 with either alone); demonstrates temporal features are most discriminative overall.