Skip to content

Rumor Cascade Models

Rumor cascade models are mathematical and computational frameworks for understanding how information spreads through social networks. They combine network structure (who is connected to whom) with individual-level adoption behavior to predict collective diffusion patterns.

Classical cascade approaches:

  • Independent cascade model: Each newly infected node independently attempts to activate its uninfected neighbors with a fixed probability; only one attempt per edge.
  • Linear threshold model: A node adopts when the fraction of its neighbors that have adopted exceeds a node-specific threshold.
  • Epidemiological models (SIR/SIS): Borrowed from disease modeling; nodes transition between susceptible, infected (spreading), and recovered (immune) states.
  • Daley-Kendall model: Early formal model of rumor spread, incorporating "stiflers" (those who learn but don't spread further).

Extensions relevant to misinformation:

  • Competing cascades: Model simultaneous spread of multiple, competing rumors or narratives rather than a single piece of information.
  • Heterogeneous spreading rates: Different rumors spread at different rates; some faster, others slower, depending on content or source credibility.
  • Directed graphs: Capture asymmetric influence (followers, citations, retweets).
  • Temporal dynamics: Include time-varying adoption probabilities, user attention constraints, and memory decay.

Modern cascade models increasingly integrate: - Cognitive factors: How biases, prior beliefs, and information credibility assessment affect adoption. - Behavioral heterogeneity: Different users have different susceptibilities, willingness to share, and engagement patterns. - Explicit competing information: Modeling simultaneous spread of true and false narratives, each influencing the other.

Key papers