Evolution of cooperation in a finite homogeneous graph

This study bridges two theoretical approaches to model the evolution of cooperation: inclusive fitness models and evolutionary game theory. Simple analytical conditions are found for the evolution of cooperation for a large class of graphs. Recent theoretical studies of selection in finite structured populations 1 , 2 , 3 , 4 , 5 , 6 , 7 have worked with one of two measures of selective advantage of an allele: fixation probability and inclusive fitness. Each approach has its own analytical strengths, but given certain assumptions they provide equivalent results 1 . In most instances the structure of the population can be specified by a network of nodes connected by edges (that is, a graph) 8 , 9 , 10 , and much of the work here has focused on a continuous-time model of evolution, first described by ref. 11 . Working in this context, we provide an inclusive fitness analysis to derive a surprisingly simple analytical condition for the selective advantage of a cooperative allele in any graph for which the structure satisfies a general symmetry condition (‘bi-transitivity’). Our results hold for a broad class of population structures, including most of those analysed previously, as well as some for which a direct calculation of fixation probability has appeared intractable. Notably, under some forms of population regulation, the ability of a cooperative allele to invade is seen to be independent of the nature of population structure (and in particular of how game partnerships are specified) and is identical to that for an unstructured population. For other types of population regulation our results reveal that cooperation can invade if players choose partners along relatively ‘high-weight’ edges.