Information cascades: the domino effect in decisions

When the first two people agree, the third person stops thinking. So does everyone after that.

Person one said yes. Person two said yes. By person five, nobody was evaluating the decision anymore — they were just joining the line.

“An information cascade doesn't require agreement. It only requires the appearance of it.”

Definition

Information Cascade

An information cascade occurs when individuals, observing the actions of others before them, rationally decide to follow the crowd and ignore their own private information, leading to fragile group consensus that can be based on very little actual evidence.

What information cascades really mean

Information cascades are subtly different from peer pressure or conformity. In a cascade, following the crowd is individually rational — each person reasons that the people before them probably had good reasons for their choice, so the accumulated social evidence outweighs their own private signal. The problem is that this logic, when applied by everyone, means the group's collective choice may be driven by just one or two early movers. Bikhchandani, Hirshleifer, and Welch formalized this in 1992, showing that cascades can start from surprisingly little initial evidence. In organizational settings, sequential decision processes — going around the table, reviewing options in order, or collecting votes publicly — are cascade machines. The order of input becomes more important than the quality of input.

Information cascades in action

The early investment rounds of WeWork demonstrate a classic cascade. SoftBank's massive investment created a powerful signal that attracted subsequent investors who performed less independent due diligence because the social proof from SoftBank's commitment seemed to provide sufficient validation. Each subsequent investor's participation reinforced the cascade, making it increasingly difficult for any individual actor to break from the pattern. When the cascade finally broke during the IPO process — when institutional investors performed independent fundamental analysis — the abrupt reversal revealed how little independent evaluation had been happening all along.

What research tells us

Bikhchandani, Hirshleifer, and Welch's 1992 model showed that cascades are mathematically inevitable in sequential decision-making with observable choices. Anderson and Holt's 1997 experiments confirmed the theory: participants reliably abandoned their private information to follow observed choices, even when their own signals were strong. Importantly, cascades are fragile — a small piece of public contradictory information can shatter them, leading to rapid reversals. Banerjee's parallel work on herd behavior showed the same dynamics in economic contexts. The key finding is that the problem isn't individual irrationality but structural: the sequential, public nature of the decision process makes cascades a predictable outcome.

Recognizing the signs

Going around the table produces monotonically increasing agreement with the first speaker
People change their stated position after hearing the majority view
Private polls produce dramatically different results than public discussions
Early momentum on an option becomes self-reinforcing and nearly impossible to redirect
New team members adopt the group's existing positions without independent evaluation
When one person finally dissents, others quickly reveal they had similar doubts all along

How to counteract information cascades

The primary structural defense is simultaneous rather than sequential input. When people commit to their views before seeing others' views — through written submissions, anonymous polls, or independent pre-meeting analysis — cascades cannot form. If sequential input is necessary, randomize the order of contribution and make it safe to break from emerging patterns. Explicitly valuing dissent, especially early dissent, raises the social reward for independent thinking. Leaders can also deliberately inject contradictory information or perspectives to test whether the apparent consensus is genuine or cascade-driven. The fragility of cascades is actually good news: because they're based on thin evidence, even a small structural intervention can prevent or break them.

78%

of participants followed the cascade even when their private information contradicted it

Source: Anderson & Holt, 1997

sequential agreements is typically enough to trigger a cascade that overrides private information

Source: Bikhchandani et al., 1992

90%

reduction in cascade formation when input is collected simultaneously vs. sequentially

Source: Lorenz et al., 2011

Frequently asked questions

Are information cascades always wrong?

No. Sometimes the early movers have genuinely good information, and following them is rational and leads to correct outcomes. The problem is that from the outside, a cascade based on good information looks identical to one based on thin evidence. The fragility is the real issue — even correct cascades can reverse abruptly when new information arrives.

How are information cascades different from groupthink?

Groupthink is driven by the desire for social harmony — people suppress dissent to maintain group cohesion. Information cascades are driven by rational inference — people follow others because they assume others have relevant information. Groupthink requires social pressure; cascades only require observable sequential choices.

Can technology prevent information cascades?

Technology can help by enabling simultaneous rather than sequential input. Tools that collect perspectives independently before revealing group patterns specifically target the sequential observation mechanism that drives cascades. The key architectural choice is whether the tool shows you others' inputs before or after you submit your own.

Ready to make better decisions?

Start your first AI-moderated session in under a minute. No meetings required.

Get Started Free