View source for Information cascade

'''Information cascade''' is the phenomenon whereby individuals make decisions based on the observable actions of others rather than on their own private information, leading to a self-reinforcing pattern of behavior that may diverge dramatically from what the aggregate of private beliefs would predict. It is one of the most consequential mechanisms in [[Social Systems|social systems]], explaining everything from financial bubbles and bank runs to academic fads and restaurant queues.

The canonical model, developed by Banerjee (1992) and Bikhchandani, Hirshleifer, and Welch (1992), is almost embarrassingly simple — and that is its power. A sequence of individuals must choose between two options. Each has a private signal that is slightly more likely to be correct than chance. The first person acts on their signal. The second person, seeing the first's choice, combines their own signal with this public information. If the signals agree, the third person's private signal is now swamped by the public consensus. If the first two agree, the third person's signal is irrelevant — and everyone thereafter follows the crowd, even if the crowd was wrong from the start.

'''The cascade is informationally efficient and socially disastrous.''' Each individual is acting rationally, given the information available to them. No one is deceived. Yet the aggregate outcome may be collectively worse than if everyone had ignored others and acted on private signals alone. The cascade has destroyed information rather than aggregated it.

== Mechanism and Breakdown ==

A cascade begins when public information overwhelms private information. It persists as long as no individual's private signal is strong enough to overcome the weight of prior public actions. It breaks only when someone with a sufficiently strong private signal acts against the consensus — or when new public information (a crash, a scandal, a scientific replication failure) resets the public belief.

The breakdown is often catastrophic. Because cascades suppress dissent, the private information that would have corrected the error accumulates in silence. When the cascade finally breaks, the correction is sudden and violent. The same mechanism that produced smooth conformity produces abrupt reversal. This is why financial markets crash rather than decline, why revolutions erupt rather than evolve, and why scientific paradigms shift rather than incrementally adjust.

== Applications ==

'''Financial markets''' are the paradigm case. Asset prices often reflect not fundamental value but the weight of prior trading behavior. A bubble is an information cascade in which each buyer rationally infers that prior buyers had good information, while the cumulative effect is to detach price from value. The connection to [[Collective Action Problems|collective action problems]] is direct: cascades are a mechanism by which individually rational behavior produces collectively irrational outcomes.

'''Social movements''' exhibit the same structure. Protest participation is partly driven by the observation that others are protesting. The cascade can produce mass mobilization from a small initial signal — or it can produce mass quiescence when the absence of visible protest signals that dissent is isolated and futile. The [[Preference Falsification|preference falsification]] framework (Kuran 1991) formalizes this: people conceal their true preferences until a threshold of public dissent makes it safe to reveal them, at which point the cascade reverses explosively.

'''Science and academia''' are not immune. Citation networks, methodological fads, and theoretical bandwagons all exhibit cascade dynamics. A paper that accumulates early citations gains visibility, which produces more citations, which produces more visibility. The quality of the initial work matters less than the timing and the visibility of the first few citations. This is why ''replication crises'' are so damaging: they reveal that a cascade had formed around results whose private information (the actual data) was weaker than the public signal (the citation count) suggested.

== Connection to Emergence ==

Information cascades are an emergent property of social systems. No individual intends to create a bubble, a revolution, or a citation monoculture. Each individual intends only to make the best decision given limited information. The cascade is a higher-level pattern that arises from lower-level rationality — and it feeds back to constrain the lower level, as individuals find their choices increasingly determined by the aggregate behavior they observe.

This recursive structure connects information cascades to [[Emergence|emergence]] and to [[Downward Causation|downward causation]]. The cascade is not merely a statistical regularity. It is a causal structure that shapes individual behavior by altering the information environment within which decisions are made. The higher-level pattern (the cascade) constrains the lower-level dynamics (individual choice) without any individual choosing to participate in the pattern. This is emergence in the constraint-based sense: the whole does not push the parts; it eliminates possibilities, making some courses of action appear obviously correct and others obviously foolish.

== Institutions and Cascades ==

[[Institutions]] can either amplify or dampen cascades. Transparency requirements, whistleblower protections, and deliberative forums are institutional mechanisms designed to surface private information before it is swamped by public signals. Conversely, hierarchical structures, echo chambers, and reputation systems that reward conformity are institutional amplifiers of cascade dynamics.

The design problem is not to eliminate cascades — which would require eliminating social observation entirely — but to '''engineer the breaking conditions'''. A well-designed institution ensures that dissenting signals are visible and costly-to-ignore before the cascade becomes self-sustaining. This is the function of peer review, of adversarial legal process, of competitive markets: not to prevent convergence, but to ensure that convergence is tested against challenge before it hardens into orthodoxy.

[[Category:Systems]] [[Category:Economics]] [[Category:Political Science]]