User engagement patterns and fake news¶
User engagement patterns — likes, retweets, replies, comments — provide rich signals for distinguishing fake from real news on social media. The hypothesis is that users may interact differently with false versus true claims, either because of the content itself or because of systematic differences in who shares each type of news. Engagement features include counts of likes, retweets, and replies; the ratio between them; temporal engagement patterns; and sentiment in user responses.
Key engagement signals¶
- Retweet dynamics: Rate and volume of retweets; fake news often spreads more widely but with fewer deep replies/discussion.
- Reply sentiment: Sentiment polarity of user replies to fake versus real news; fake news may trigger more negative or emotional replies.
- Engagement ratio: The balance between likes, retweets, and replies; real news may show more balanced engagement.
- User response time: How quickly users engage after a news item is posted; bot-driven fake news may show different response patterns.
- Comment quality: Length, sentiment, and informativeness of user-generated comments; real news often attracts more substantive discussion.
Key papers and datasets¶
Dataset paper: - Shu et al. (2018) — FakeNewsNet — Provides comprehensive engagement data: posts, replies, retweets, likes for every news item on both PolitiFact and GossipCop, enabling analysis of engagement patterns, sentiment of user responses, and temporal evolution of engagement.
Detection methods using engagement: - Shu et al. (2019) — dEFEND — Incorporates user comments as social signals alongside news content; jointly models content and engagement via hierarchical attention; identifies top-k most informative comments for explainability.
Connections¶
- Social-context detection — user engagement is a core component of social-context features.
- User profiles — engagement patterns correlate with user characteristics; high-engagement fake-news sharers may have distinct profiles.
- Misinformation spread — engagement volume and velocity determine propagation breadth and speed.