Information Cascade in Social Networks

Threshold model: agents adopt when enough neighbors have adopted

0
Adopted
0%
Fraction
0
Wave
Cascade?
Susceptible
Just adopted
Adopted (earlier)
Seed node
0.25
4
3
4
Granovetter's threshold model (1978): each agent has a threshold φ — they adopt the information/behavior only when the fraction of their neighbors who have already adopted exceeds φ. Even with identical thresholds, outcomes depend critically on network structure and seed placement.

A global cascade occurs when a small seed triggers adoption across most of the network. Watts (2002) showed cascades require: (a) a connected component of low-threshold "early adopters," and (b) seeds placed within it. The cascade window — the range of connectivity where global cascades are possible — is narrow. Too sparse: cascade dies early. Too dense: each node needs many neighbors to adopt first. Click New Network to regenerate, then Run to watch the cascade propagate.