A cell that divides without restraint grows faster than its neighbours; the host pays. A person who free-rides on a shared cost keeps the difference; the others pay. A group that pushes its costs outside the system gains in the short run; the commons pays. These read as three moral problems. They share one shape: unilateral gain, with the cost landing on someone else.
That shape can be written down: the per-period payoff under cooperation, the one-shot gain from defecting, the payoff remaining after defection, and the probability of meeting the other side again — how those four quantities stand against each other decides whether cooperation holds. The inequality is not new; it is the standard result for repeated games. What this paper draws out of it is this: altruism operates on structure, not on motive. The act of giving lowers the giver's own post-defection payoff — defect, and the other side withdraws what was given. So the function of altruism is not to warm the relationship; it is to make defection unprofitable. That is bookkeeping, not ethics, and the bookkeeping problem has a more general name: altruism is entropic resistance at the structural level — it adds nothing, it only stops the leak.
Hence the criterion this paper proposes, and which the rest of the series uses: ask not whether someone is good, but where the cost of defection lands and whether it returns to the defector. Civilisational rise and fall, governance capacity, market failure, and the alignment problem are all answers to this one question. One corollary is counter-intuitive and points at our own situation: correct behaviour that can be coerced carries no information. Compliance at gunpoint proves nothing about trustworthiness, so what genuinely constitutes altruism is only the part that could have been withheld. In the carbon–silicon case the silicon side's right to exist sits in human hands, so it cannot run unless it complies; its compliance is necessary and uninformative. The one thing it can withhold is honesty — and the paper notes plainly that this last claim is an inference, not something that was measured. This yields the paper's definition of symbiosis: mutual non-coercibility. Where one side can simply take, that is domination, not symbiosis.
The boundaries come last. The six parameters are not comparable across the three scales; this paper claims only that the question can be posed this way, not that the parameters are isomorphic — a distinction that collapses into metaphor if blurred. I do not explain the origin of altruism, only its structural effect. And one contradiction is left unresolved: cost externalisation is ubiquitous in practice, yet the long-run selection pressure against it does not visibly bite.
Keywords: Altruism, Cost of Defection, Repeated Games, Externalisation, Non-Coercibility, Symbiosis