Information Cascade 3D Simulator

Information Cascade — 3D

Private signal accuracy 70%

Sequence speed 2.0 / s

Cascade state

–

Locked on

–

Ignored own signal

–

Cascade vs truth

–

Agents decided

–

Net public votes

–

How it works

Each glowing node is one agent in a queue that spirals outward and upward, decided strictly in order. A hidden true state Θ (A or B) is fixed at the start; every agent draws a private signal that matches Θ with probability p, but the node stays grey until its turn arrives. At that moment the agent combines two log-odds terms — its own signal and the tally of everyone who chose before it — exactly as in the flat 2D version of this model:

L_self  = ±log(p / (1−p))
L_crowd = (n_A − n_B) · log(p / (1−p))
choose A if L_crowd + L_self > 0, else B

Watch the central gauge column: it rises for a lead toward A, sinks for a lead toward B, and the moment it crosses either dashed threshold ring (±2 net votes) every remaining node stops pulsing through a "considering" animation and instead snap-decides instantly, in the gauge's locked color — a visible tell that public evidence alone, not the node's own private signal, made the call. Nodes that decided this way carry a thin dashed halo ring, matching the 2D chart's convention.

Private signal accuracy (p) — reliability of each node's own evidence. The ±2-vote cascade threshold is the same for every p<1; only the odds of an early, correct lead change.
Sequence speed — how fast the spiral extends. Purely visual pacing.
Restart — redraws Θ, respins every private signal, and rebuilds the spiral from node zero.

Same Bikhchandani–Hirshleifer–Welch cascade math as the 2D version — here rendered as a physical structure you can orbit around to see the lock-in happen live.

Chose A Chose B Undecided Ignored own signal

⚠ Couldn't load the 3D engineThree.js failed to load from the CDN. Check your connection and reload.