Information Cascades

When you pick a busy restaurant over an empty one, you've experienced an Information Cascade. This happens when we copy the actions of others while ignoring our own information, assuming the crowd knows something we don't. This process can create massive trends from small beginnings, like a single video going viral. Try this: next time you see a long line, ask yourself if the item is genuinely good, or just popular because others are in line.

An information cascade begins when a few individuals make a public choice. Observers who come later see these choices and may rationally decide to ignore their own signals—like a mixed online review or personal preference—and follow the group's apparent consensus.

This creates a chain reaction. The third person sees two people made the same choice, making it even more compelling to follow suit. The fourth person sees three, and so on. At a certain point, the sheer weight of public action overpowers an individual's private knowledge.

For example, in corporate hiring, if the first two interviewers on a panel express positive feedback about a candidate, the third interviewer may downplay their own private doubts, assuming their colleagues spotted something they missed. The initial positive signal cascades through the group, leading to a consensus that might be based on weak evidence.

This phenomenon also drives fashion trends. An obscure brand worn by a few influential celebrities can trigger a cascade where thousands of people buy the item, not because of its inherent quality, but because they infer value from the actions of early adopters.

The key takeaway is that cascades aren't necessarily irrational; they are a logical shortcut. The danger is that the initial choices might have been random or wrong, leading an entire group astray.

The concept of information cascades was formally developed to explain how seemingly rational individuals could collectively participate in fads, bubbles, and mass behaviors that are not supported by strong evidence. These models show that when people can observe actions but not the private information behind them, conformity can emerge rapidly from a small initial seed.

• Bikhchandani, Hirshleifer, and Welch (1992) developed one of the first formal models, demonstrating that cascades can be triggered by the decisions of just a few individuals, leading to uniform but fragile social outcomes. [1]
• Anderson and Holt (1997) confirmed these theoretical predictions in laboratory experiments where participants would frequently ignore their own correct private signals to follow the incorrect public choices made by others before them. [2]
• Salganik, Dodds, and Watts (2006) showed in their 'MusicLab' experiment that when people could see what songs were popular, they were more likely to download those songs, suggesting that social influence can artificially create 'hits' and make success much less predictable. [3]

The pure model of an information cascade has specific requirements that aren't always met in the real world. For one, it assumes people make decisions in a clear sequence. If many people decide simultaneously, a cascade is less likely to form. The model also assumes people only see actions, not the reasoning behind them. If individuals can communicate their private knowledge, the cascade can be prevented or broken. Finally, not all herding behavior is an information cascade; sometimes it's driven by social pressure (conformity) or network effects rather than informational inference.