Group selection can roughly be defined as a form of natural selection by which the fitness (probable reproductive success) of the group is maintained or enhanced, even at the cost of reduced fitness of individuals in that group.
It is generally considered, within current or recent evolutionary biology, that group selection is an error, a misunderstanding, the explanation or last-resort or so rare in nature as to be insignificant.
By contrast, I regard group selection as necessary for any other kind of selection - in other words, without group selection there can be no build-up of complex adaptations: group selection is necessary in order to maintain complex entities - indeed group selection is necessary to the sustainability of life itself.
So, group selection - far from being insignificant or a last resort, is ubiquitous, found everywhere.
However, in arguing this I am engaging in a re-conceptualization of group selection - in particular I am using definitions derived from a particular theory of complex systems: that associated with Niklaus Luhmann.
[See Appendix and associated references in:
In considering group selection it is worth having a definition of what constitutes a group of the type capable of acting as a selection pressure on the individuals composing the group; since the minimal requirement of group selection is the presence of a suitable group that is stable (in those key respects which enforce selection) over a sufficient number of generations.
(Most of the ways of conceptualizing groups selection are, I believe inadequate)
This question can be reduced to the definition of a biological entity.
A biological entity is a (relatively) concentrated and sustained network of communications and interactions. Thus a single cell, a multicellular organism, an organ (such as the heart, or a gland) within a multicellular organism, and the social group in social organisms may all be considered to be entities.
For example, in social animals (including humans) that which makes them social animals can be defined as a sufficient density of particular types of communications and interactions between individuals - sustained across sufficient numbers of generations - such that this acts upon the individuals to shape behaviour by mechanisms that are transmissible between generations (for example, by inducing genetic, or some types of epigenetic, change of individuals).
In other words, the group itself displays a system-autonomy from the individuals which compose the group - therefore the group itself is a selection pressure on the individual animals within the group.
What is meant by autonomy? That the group entity operates to sustain and expand and/or reproduce itself - this being a defining property of all complex systems.
If an emergent complex system (occurring by chance) were to lack the ability to sustain and reproduce itself, it will soon simply cease to be - and would not be observable (or only momentarily so).
This is because a system is defined in contrast to its environment, and the basic property of a system is to separate itself from the environment in a context where chance/ entropic change will tend to assimilate the system into the environment (e.g. this happens after death).
Therefore any complex system which is sustained, must have the ability to sustain itself - to re-make and re-produce itself in the context of its environment - and for social groups individual organisms are a part of its environment. Therefore, social groups are buffered against the individuals which constitute it - but this is not a paradox, because what makes the social group definable as a complex system is the interaction of inter-individual communications, therefore not the actual physical individual organisms.
This may clarify the current confusion about what is 'doing the selecting' versus what 'gets selected' - the so-called levels of selection problem (with levels such as genes, organisms or groups). Instead, of physical things, the problem is reconceptualized in terms of communications; instead of physical things, the problem is reconceptualised in terms of abstract systems; instead of hereditary information conceptualized in terms of physical things, the problem is reconceptualised in terms of systems having the intrinsic property of self-sustaining and reproduction.
Instead of looking at things such as gene frequencies in populations, the focus moves to things like the density, frequency and complexity of communications and interactions.
This, of course, represents a 'paradigm shift' in discussing selection in biology - and it is notoriously difficult to argue in favour of a paradigm shift - especially when the current paradigm is yielding security, status and funding.
Nonetheless, if the new complex-systems paradigm can be grasped, I think its explanatory superiority - especially in terms of clarity, and in terms of explaining some of the key problems of biology such as 1. the origins of biological life and 2. the major transitions in evolution (e.g. evolution of the cell, the eukaryotic cell, multicellular organisms, sexual reproduction, social organisms) is very clear.