Step 0

Idea Spreading on a Friendship Graph (2D)

How a harmless idea or habit travels through 60 friends. Twin of the 3D ring of people.
Switch to 3D
Heard it so far-
Sharing now-
Links-
Mean friends (this graph)-
R0 = T x k-
Final share (theory z-squared)-
Share of people by step: not heard (grey), sharing (orange), heard and stopped (green)
Final share vs chance T: theory z^2 (violet), simulation on 1500-person random graphs (yellow), threshold 1/k (red)
Friends per person on this network (blue dots) vs Poisson with the same mean (yellow dashed)
The science
A teaching model of information or a harmless habit spreading. It says nothing about any real group, and nothing about people's health or beliefs.

Spreading on a network. Each of the 60 people is not heard (S), sharing (I) or heard and stopped (R). In every step each sharing person tells each friend who has not heard it with probability T and then stops sharing. This is the SIR model on a graph; for this one-shot rule it is exactly bond percolation: every friendship link independently "works" with probability T, and the idea reaches everyone connected to the starting person by working links.

The threshold and the final size. On a large random graph with mean degree k the basic reproduction number is R0 = T k. Below R0 = 1 the idea dies out; above it a share z of people is reached, where z = 1 - exp(-R0 z). A single random starter lands in that giant group with probability z, so the expected final share is about z^2 (violet curve; yellow dots are simulations).

New in this 2D view. The raster at the bottom lists people in the order they heard the idea; the diagonal edge is the wave front. The layout is a force-directed drawing (links pull, people push apart), so clusters and shortcuts are visible, unlike the fixed ring of the 3D page. Two extra checks confirm the percolation picture directly.

Simplifications: real people hear ideas more than once, forget, doubt, and are influenced by several friends at once (complex contagion); friendships are not equally strong and not fixed; time is discrete and everyone acts in step. 60 people is small, so single runs are noisy and differ from the large-network formula: that is why check 1 uses 3000 people.

Independent numerical checks
Made with 🟠 Blender