Figure 2. Frequency of cooperation achieved in a Prisoner's Dilemma game as a function of the benefit of cooperation and the incongruence between the interaction and the external learning network.

The simulations are based on scenario 2. Both networks can either be random-regular or scale-free. Cooperation is strongly favoured when the interaction network is scale-free (bottom row) and weakly favoured when the external learning network is scale-free (right column).