The article “Scale-Free Networks Provide a Unifying Framework for the Emergence of Cooperation” by F.D. Santos and J.M. Pacheco suggests that structure of a network influences which strategies evolve within a population. Two games one of which is familiar to our class – the prisoner’s dilemma and a variation called “the snowdrift game”- are used to examine how cooperative strategies might spread on a network.

On a large scale, network cooperation is not an evolutionarily stable equilibrium because it is vulnerable to invasion by defector strategies (R < T). The Snowdrift game is more favorable to cooperation as it removes defection as the dominant strategy. Similar to how neighbors influence an individual during diffusion of a network, this study relied on neighbors’ payoffs to determine whether an individual would switch strategies.

Previous studies found that the role of spatially structured populations (much like the ones we have been studying in class) on the emergence of cooperation varied from game to game. In contrast, this study actually found that when networks were grown using Barabasi’s preferential attachment model (described previously and in class) cooperation becomes a predominant trait for both games on both large and small networks at equilibrium independent of how the relative payoffs are set. The explanation proposed that hubs within the network are havens for cooperators. As the hubs are highly interconnected, they resist invasion by lone defectors. They found that if they removed the age-correlations that Barabasi’s model encourages then cooperation is less prevalent.

This article suggests that network structure is a necessary topic of conversation alongside whether a strategy is evolutionarily stable. I would agree as the evolutionarily stable strategy discussion harshly requires an entire population to pick one strategy that is robust against any attack. This seems to overlook some of the dynamics we discussed pertaining to information cascades and the role of clustering. If network structure influences what strategies survive, could we design networks (or the way they grow) so that over time social welfare maximizing strategies were picked over individual welfare maximizing strategies?