Altruism pays off in group-structured populations through probable reciprocity

The evolution of altruistic traits is a Darwinian puzzle that has not yet been satisfactorily solved. Using multi-agent simulations for public goods games and a threshold model, I show here that there is a clear fitness advantage for the altruists in a mixed population of altruists and egoists if the following conditions are met: (i) The population is segregated into groups that persist over time or several generations. (ii) The benefit achieved by altruists is only shared among the members of their own group. (iii) There is selection pressure on each individual, which is reduced by the benefit provided by altruists. (iv) The costs for the altruists are smaller than the overall benefit they achieve for all members of their group together. (v) There is already a high proportion of altruists in the population and/or the proportion of altruists varies greatly between groups. I refer this kind of population dynamics as the grouping effect, or G-effect for short. This effect is made up of two biases, which basically come about through simple individual selection: the tendency to increase the variance among the groups (variation bias) and the tendency to increase fitness for altruistic traits when this variance is high (fitness bias). In this context, selection only happens at the individual level. It does not require relationships between the group members, nor does it require competition among the groups. Extended multi-agent simulations with reproduction, mutation and migration also show that altruism can exist stably in a wide range of conditions due to this G-effect and the resulting probable reciprocity.