Rational agents sequentially decide — private signals discarded once cascade forms
The information cascade model (Bikhchandani, Hirshleifer, Welch 1992) shows how rational Bayesian agents can collectively make the wrong choice. Each agent receives a private signal about the true state (A or B) with accuracy p, observes all previous public decisions, and acts on the posterior belief. Once two consecutive agents choose the same action, the public belief dominates — subsequent agents rationally ignore their private signal and herd, even if the cascade chose the wrong state. Social learning becomes fragile: a single early run of bad luck can lock in a wrong cascade permanently.
Bayesiansocial learningherdinginformation cascades