The Synergistic Threshold

Complex contagion theory predicts that some behaviors spread only when people see multiple independent sources of social reinforcement. One friend telling you to try something is ignorable. Two friends telling you independently is different — not twice as persuasive, but qualitatively more persuasive. The second signal changes the interpretation of the first.

The theory has been debated for two decades. Lab experiments support it. Observational studies are confounded — people with two adopting friends differ systematically from people with one. The causal question requires randomized exposure to exactly one or exactly two independent sources of influence, at scale, in a natural social context.

This paper ran the experiment. A country-scale field trial randomly assigned individuals to receive encouragement from either one or two friends to share a mobile data coupon. The design is clean: the number of encouraging friends is randomized, the behavior is measurable (coupon sharing), and the social network is known.

Complex contagion works. Individuals exposed to two friends adopted at significantly higher rates than those exposed to one. The effect is not additive — the second friend's encouragement doesn't simply add a fixed increment. The signals interact synergistically. Two sources of social reinforcement produce more adoption than two independent cascades would predict.

Network embeddedness moderates the effect. When the two encouraging friends are themselves connected — when they form a closed triad rather than independent paths — the reinforcement is stronger. The structure of the network around the target matters as much as the number of signals reaching them.

The through-claim: social influence is not a force applied to an individual. It is a property of the configuration around the individual. One signal is information. Two signals from independent sources is social proof. Two signals from connected sources is a norm. The same person receiving the same message changes their behavior based on the topology of who sent it. The message didn't change. The network around it did.