Dynamic Social Networks Promote Cooperation in Experiments with Humans

Investigators set out to test the theory that people change their behavior in response to their social partners. They conducted experiments using an artificial social network and online economic game. In the game, each subject interacts with his/her neighbors by either cooperating or defecting. When cooperating, individuals pay 50 units to each neighbor, resulting in the neighbor gaining 100 units. Defection results in no costs or benefits to either party. Before making each decision, subjects are reminded of their neighbor’s previous decision.

The investigators examined three kinds of network conditions:

  1. Random link—20 percent of possible links are randomly updated to create a well-mixed population.
  2. Fixed link—the network is static and remains so.
  3. Strategic link—a percentage of pairs are randomly picked to have their connections updated (kept or broken), being aware of the other’s action in the proceeding round.

In one of two strategic link scenarios, the network updates infrequently, replacing 10 percent of pairings each round (viscous). In another, the network updates more frequently, with 30 percent of pairings potentially changing each round (fluid).

As expected, investigators found cooperation declines steeply over time in the random and fixed network conditions. In the viscous network, cooperation also decreases over time. However, in the fluid network cooperation remained robust and stable throughout the experiment. The findings from the fluid network reinforce the fact that when subjects can preferentially break social links with defectors and maintain connections with cooperators, an incentive to cooperate is created and network structures can be maintained.