Information Cascades

Theoretical and empirical study of how sequences of individual decisions aggregate information. A cascade occurs when users ignore their own private signals and follow earlier decisions, either because they assume others have superior information or because public agreement creates strong social pressure.

Cascades can be either informative cascades (the crowd has converged on the correct answer) or uninformative cascades (early errors lock in and persist, even as later evidence contradicts them). In the misinformation setting, cascades are especially dangerous: false rumors, fabricated evidence, and misleading framing can cascade through networks of millions of users before fact-checks reach them.

Recent work has studied how platform interventions (credibility labels, expert review, algorithmic diversification) and AI-based signals reshape cascade dynamics, either preventing false cascades or accelerating correction.

Key papers

• One AI Signal, Many Human Judgments: A Bayesian Cascade Analysis of AI-based Credibility Indicators in Online Information Spread — extends classical cascade models to include AI credibility indicators as shared public signals
• A Survey of Information Cascade Analysis: Models, Predictions, and Recent — comprehensive survey of cascade prediction and modeling

Related topics

• Social learning and belief transmission (mechanisms of learning from others)
• Rumor detection on social media (identifying cascading rumors)
• Network science (graph-based cascade analysis)
• Fact-checking and corrections (interventions to stop false cascades)