Close-knit neighborhoods: Stability of cooperation in networks

Agents connected in a network face a binary choice whether to contribute or to free-ride. The former action is costly but benefits the agent and her neighbors, while the latter is free, but does not provide any benefits. Who will contribute if agents are farsighted and not constrained by a fixed non-cooperative protocol? I adapt the concepts of consistent sets and farsightedly stable sets to answer this question. When benefits to an agent are linear in the number of her contributing neighbors, the decision to contribute depends on the cohesion of her neighborhood as captured by the graph-theoretical concept of k-cores.