Why does anything help anything else?

A 2006 review set out five mechanisms and the inequality each one requires. Each describes a different circumstance, and more than one can operate at once: vampire bats share food with relatives, but what a bat has been given predicts what it gives better than relatedness does.

What they share is that cooperation is never free. Each specifies what has to be true about relatedness, repetition, reputation, space or group structure for the cost to be repaid.

J. B. S. Haldane anticipated the first of the five, and W. D. Hamilton wrote it down in 1964 as an inequality: help a relative when the benefit to them, multiplied by how closely related you are, exceeds the cost to you.

It reframed the unit of accounting. A gene that causes its bearer to help relatives is helping individuals likely to carry copies of it, so behaviour that looks costly to an individual can be profitable to the gene producing it.

A second paper the same year turned the idea on the worker castes of the social insects. Because of how sex is determined in the Hymenoptera, a female whose mother mated only once is more closely related to her sisters than she would be to her own daughters, so raising the queen’s brood can pay better than laying. Hamilton put that forward as a hypothesis and said the evidence then available was somewhat contrary to it on certain points.

It has not held up. In an exchange in Nature in 2010 and 2011, critics of kin selection theory and 137 of its defenders agreed on one point: haplodiploidy by itself probably has only a minor bearing on how worker castes arose. The defenders point instead to monogamy: worker castes, they write, evolved only in monogamous lineages.

What the inequality did settle was the standing of the question. It stopped being an objection to natural selection and became a subject within it.

Genes cooperating in a genome, cells cooperating in a body, individuals cooperating in a colony: each is a larger unit built from smaller ones that forgo some of their own reproduction to help one another.

Cancer is what happens when that settlement breaks down in one of them, which is a useful reminder that the cooperation inside a body is enforced rather than automatic.

Humans cooperate with strangers they will never meet again, in experiments where reputation is controlled away. On their face, kin selection and direct reciprocity do not cover that.

Candidate accounts include a disposition to reciprocate that evolved where a one-off encounter was hard to tell from a repeated one, norms and institutions built up over human history along with the punishment that enforces them, and competition between groups. The argument among them is active. The experiments are also hard to read: participants know they are being watched, and Levitt and List argue that this exaggerates how prosocial they appear.

Whether selection acting on groups explains anything that kin selection does not is one of the long-running disputes in evolutionary biology, and it is partly a dispute about accounting rather than about biology. Both sides have generally accepted that the two descriptions are formally equivalent, so the same outcome can be told either way. Van Veelen argued in 2020 that this fails where the effect of cooperating depends on how many others cooperate.

Where it is not merely bookkeeping is in the empirical question of whether the conditions for group selection are met often in nature, and that is not settled.