Blast from the recent past. This is pertinent to the previous discussions on evolution as a special case of emergent-system emergence.
It was argued that "group selection" effects have been discredited in evolutionary biology. I counterargued that denying the possibility of a selection effect at each more-and-more complex system-level in a multi-layer complex-ordered emergent system (such as ecosystems, biological species etc) denies the likelihood of spontaneous emergence of those complex systems at all.
I think I've found the source of the confusion regarding group selection effects. It goes like this:
A species can evolve a "group-benefit" behaviour so long as the development
of the behaviour does not, on average, reduce the reproductive success of individuals
that engage in the group-benefit behaviour, and so long as the behaviour does
confer, on average, a benefit to the reproductive chances of each individual in
the well-behaving group.
The key is in how we interpret "average". The question is whether an individual
organism always acts "in each short-term encounter" in a manner which maximizes their
chance of survival-to-breeding-age IN THAT ENCOUNTER, or whether it is possible
for the individual to wager that taking a slight risk now (and believing or observing that
others will also do so) will lead to a better chance that the individual will survive ALL
ENCOUNTERS from now up until it breeds. The organism doesn't have to be smart enough
to believe in this wager. It is sufficient that the wager be on average beneficial to the
individual.In that case, through repeated trials by multiple individuals, the behaviour
which is group-adaptive and individually "lifetime-average" adaptive can evolve.
BECAUSE THE EVOLVABLE "GOAL" IS NOT SIMPLY TO MAXIMIZE THE CHANCE OF SURVIVAL OF AN ORGANISM OF THE NEXT SHORT-TERM ENCOUNTER. THE "GOAL" IS TO MAXIMIZE THE PROBABILITY OF SURVIVAL OF THE SUM TOTAL OF ALL OF THE ORGANISM'S ENCOUNTERS UP TO WHEN THE ORGANISM REPRODUCES.
So it is just a time-scale misunderstanding. Group-adaptive behaviours increase the member's
probability of surviving to reproductive age, even if they slightly increase the chance of the
indvidual losing some particular encounter.
True "extreme" altruistic behavior which conveys CERTAINTY of death in a single encounter
may not fit into this model, but it can be argued as to whether the altruistic individual "believes"
they are going to die "for certain" in many incidents or not, or whether they hold out "faint hope"
in which case the argument above could still hold. In any case, true "certain death" altruistic behaviour
is an extreme anomoly case of group-adaptive behviour. Most group-adaptive behaviours are
not of that kind, so "extreme, definitely fatal" altruism is not a good model for them.