There are a number of reasons why this is true. It is simple, and therefore, users can apply its logic with ease; nevertheless, it is quite general. In some versions, it is exact, and even less exact versions are not necessarily a strong concern for field or comparative studies, where we can only measure crudely anyway. Crucially, it is often sufficiently independent of the genetic details, such as dominance and recessiveness, the number of genes, and their allele frequencies. This allows it to become an important tool of the phenotypic gambit (Grafen, 1984) and optimality approaches. It can be used for traits where we do not understand the underlying genetics, and, in fact, we never fully understand the genetics. It also conveniently separates selection into two kinds of summary terms: effects on fitness (costs and benefits) and population structure (relatedness). This separation makes the process easy to think about and the equations easy to apply. Inclusive fitness points to cause-effect relations, specifically to the various effects caused by the actor’s behavior. This focus on what the actor can control allows us to tie into the long biological tradition of thinking of actors, or their genes, as agents. Additionally, it tells us that these agents should appear to be trying to maximize inclusive fitness.
Inclusive fitness is not perfect. It does not provide the most natural way to handle explicit dynamics. It usually takes population structure as a given, and when it does this, it may not yield insight into how population structure emerges. Although, in principle, it covers everything, its summary parameters may sometimes conceal interesting complexity. Even its treatment of social causation is incomplete. For example, although it would include any benefits from mutualism in with other effects on the actor’s direct fitness, it does not usually separate out these effects or provide a causal treatment of them. Many or all of these deficits are fixable, although sometimes at the cost of making the models more complex and therefore, losing some of the advantages of the approach. In this paper, I will try to expand the types of social causation covered explicitly, while trying to maintain reasonable simplicity. For example, I will show how to specify mutualistic social effects in a category that I call kith selection, named after the largely archaic word for acquaintances, friends, and neighbors.
I will also argue that it is often worth distinguishing kin and kith selection from what I call kind selection, partly to properly capture social causality and partly because these forms of social selection act in very different ways. Inclusive fitness, developed by Hamilton (1964a), is closely associated with the process of kin selection, named by Maynard Smith (1964). However, they are not the same thing. Inclusive fitness is an accounting method and maximand. Kin selection is a process, and it can be described by other kinds of accounting. The obvious example is the neighbor-modulated approach that uses the same fitness partition as