Kin selection vs multilevel selection: what is the debate?

Kin selection and multilevel (group) selection are two frameworks for explaining how costly cooperative or altruistic behaviour can evolve. Most biologists regard them as formally equivalent, but a growing number of dissenters do not. The argument is now less about whether either framework can compute evolutionary change and more about what the mathematical equivalence shows about causes and explanation. [1] [2]

Both approaches need positive assortment: genetically similar individuals must interact with one another more often than chance would predict. In the multilevel view, assortment makes groups internally more similar, which increases the scope for selection between groups relative to selection within them. In the kin selection view, assortment directs altruism toward individuals who also carry the altruistic trait, raising the altruist's inclusive fitness above that of non-altruists. [1]

A review in the Stanford Encyclopedia of Philosophy also notes a difference in research focus. Kin selection work tends to aim at identifying character states and genic or kin-based processes that maximize fitness, while multilevel and group selection methods are oriented differently. [3]

The usual argument uses the Price equation, which expresses evolutionary change as a covariance between a trait and fitness. It can be partitioned into direct and indirect effects, from which Hamilton's rule is derived, or into individual-level and group-level effects. Both partitions correctly compute the total change, so they are seen as complementary ways of viewing evolution in structured populations. Two caveats apply: the multilevel partition covers only one type of scenario (MLS-1) and requires individuals to sit in non-overlapping groups, while kin selection has no such requirement and is arguably more general. [1]

One line of criticism says most critics contrast narrow versions of kin selection with broader multilevel approaches. On that view kin selection is a special case of multilevel selection, because genetic similarity can arise without kinship, as with hypothetical green-beard traits. [1]

A different objection, in Current Biology, calls the equivalence thesis not a deep biological result. The author argues it comes from stripping both theories of most of their biological content and noting that two abstract statistical formalisms match. In this view, the terms should label causal hypotheses about how social behaviour evolves. [4]

Methods are contested too. Contextual analysis, common in empirical work, detects group selection only if group effects remain after controlling for individual effects. The Price partition can find a group component even in non-social cases. Contextual analysis has its own problems, so it remains controversial which is better. [1]

A systematic review of 94 model-centered studies found that among the 68 that substantively discussed equivalence, qualified positions predominated and categorical rejection was rare. Its authors conclude that formal equivalence can link representations without showing they identify the same causal structure or explanatory function. Other authors suggest a bipartisan view, in which one framework may suit a given causal situation better even if the two are equivalent statistically. They add that the debate is partly philosophical and cannot be settled by theory or data alone. [2] [1]