User behavior

Analysis of how users interact with and promote fake news, including engagement patterns and group behavior.

Key papers and articles

• Dou et al. (2021) — User Preference-aware Fake News Detection (UPFD): Encodes user endogenous preferences from historical social media posts (recent tweets) to model confirmation bias and news sharing behavior. Combines user preference embeddings with news propagation graphs via GNNs; demonstrates that user historical behavior is more predictive than profile features alone (BERT+GraphSAGE: 84.62% accuracy on Politifact, 97.23% on Gossipcop).
• Tacchini et al. (2017) — Some Like it Hoax — Exploits user interaction patterns (likes) to identify hoaxes with >99% accuracy. Users preferentially like certain types of content; the set of users who liked a post is highly predictive of whether it is a hoax.
• Ruchansky, Seo, & Liu (2017) — CSI: A Hybrid Deep Model for Fake News Detection: Models user suspiciousness through co-engagement patterns on articles; identifies users who coordinate promotion of fake content through group behavior analysis.