Cascades model

v0_41
Walkthrough parameters & run controls
Private evidence: A Private evidence: B Public vote: A Public vote: B
True state—Model constants—
Single community demonstrations & aggregate metrics
Communities (same signals, different τ; sorted high→low)
—
No-cascade & Incorrect Cascade & Incorrect No-cascade & Correct Cascade & Correct
Event log
L = log-odds; λ = log(q/(1−q))
Parameters & batch controls
Main τ figure
Order parameters over τ
Empirical P(cascade) Empirical P(correct aggregate vote) Predicted phase P(cascade) Predicted phase P(correct aggregate vote) Phase cutoff τ = 1/d
Phase performance
τ phase performance
Stack grouped by

Each τ row uses the same simulated communities. Within a row, column N classifies each community using only its first N agents.

3D interactive view

N and τ form the horizontal plane; height shows either stacked outcome share or an order parameter, depending on the selected view.

Run the τ sweep to generate the interactive 3D view.
Drag to orbit, Shift-drag to pan, and scroll to zoom. Click a legend item to hide or restore a trace; double-click one to isolate it. In volume mode, each colored cuboid is a mutually exclusive outcome slice inside a full N × τ column. In order-parameter mode, the two surfaces show P(cascade) and P(correct aggregate vote).