In The Righteous Mind, Jonathan Haidt pushes hard on the hypothesis that mechanisms of coordination have been selected for at group level. He doesn’t come down as saying it definitively, but he does summarise his position as:

Natural selection works at multiple levels simultaneously, sometimes including groups of organisms. I can’t say for sure that human nature was shaped by group selection—there are scientists whose views I respect on both sides of the debate. But as a psychologist studying morality, I can say that multilevel selection would go a long way toward explaining why people are simultaneously so selfish and so groupish.

It’s only the genuinely altruistic aspects of coordination that need to be explained by group selection. If the benefit to coordination is enjoyed by all parties then you don’t need to invoke the better performance of the group in order to explain its evolution. So Haidt points to some things that do look genuinely altruistic.

But there’s a question that lingers in my mind: Having discovered that Zahavi’s handicap principle operates in status games and that people use ostensibly altruistic behaviour to indicate fitness, how can you disentangle genuinely altruistic behaviour explained by a group-level selection advantage from ostensibly altruistic fitness displays with an individual-level selection advantage?

I can think of a couple of possibilities: Firstly, if the altruistic act is done without other people knowing, then it’s much less likely to be the result of the handicap principle. Secondly, if the act is group-dependent, i.e. you would do it for a member of your group but not a member of a rival group, then it’s less likely to be a result of the handicap principle. The handicap principle should be equally applicable for any altruistic act.

I’m also not even sure how to think about cases of clear self-sacrifice, e.g. going into battle. Depending on the gains from acts of heroism, and depending on whether defectors are punished, some risk of self-sacrifice may be worth taking.