Cooperation among unrelated individuals should be evolutionarily impossible. If you help someone and they don't repay you, you've wasted energy on a non-relative who doesn't share your genes. The defector wins. In a world of non-cooperators, the first person to cooperate loses.
Yet animals cooperate constantly without genetic kinship. Fish school together for protection. Birds fly in V-formation, taking turns in the energy-draining front position. Vampire bats feed each other's babies. Unrelated primates groom, mob predators, and share meat. How?
The answer is reciprocal altruism: "I'll scratch your back if you scratch mine. I'd rather not actually scratch yours if I can get away with it. And I'm watching you in case you try the same."1
The system requires: social contact frequent enough that cooperators encounter each other again, individual recognition so you know who owes you, and a capacity to remember who cheated. Once those conditions exist, cooperation can evolve.
To understand when cooperation is rational, game theorists use the Prisoner's Dilemma (PD). Two gang members are arrested separately. Prosecutors lack evidence for the major crime but can convict them on a lesser charge (one year each). Each prisoner is offered a deal: inform on the other and your sentence is reduced.
The outcomes:
Rational analysis: If you cooperate and they defect, you get 3 years. If you defect and they cooperate, you go free. If both cooperate, you get 1 year. If both defect, you get 2 years. The rational move is always to defect — you maximize your payoff regardless of what they do.
But here's the trap: if both follow this logic, both defect and both get 2 years, when they could have both stayed silent and gotten 1 year. Rational individual behavior produces irrational collective outcome.
In single-round PD, defection is always optimal. But what if the game repeats indefinitely?
Political scientist Robert Axelrod asked game theorists what strategy they'd use in an infinitely-repeated PD. The answers were complex and varied. Axelrod programmed them and ran a tournament.
The winner was the simplest strategy: Tit-for-Tat. Cooperate on the first round. Then do whatever your opponent did last round. If they cooperate, you cooperate. If they defect, you punish them once, then resume cooperation if they do.
Tit-for-Tat succeeds because it's nice (starts cooperative), retaliatory (punishes defectors immediately), forgiving (resumes cooperation), and clear (opponents can predict your behavior and coordinate). The strategy is optimal not by being aggressive or exploitative, but by being reliable and fair.
Humans and other animals use Tit-for-Tat intuitively. We keep track of who helps us and who betrays us. We reward cooperation and punish defection. We forgive those who defect but resume cooperating. This is why reputation matters so much — your past behavior predicts your future behavior, and other people are watching.
From a behavioral-mechanics perspective, reciprocal altruism is a system of contingent cooperation: "I help you because I expect you to help me later." This creates leverage. You can control someone's behavior by:
This is why small communities (where everyone knows everyone's history) have low crime and high cooperation, while anonymous contexts (where reputation doesn't travel) have high defection. The threat of future non-cooperation — social exclusion, reciprocal non-help, reputational damage — is the mechanism that sustains cooperation.
Behavioral-mechanics describes reciprocal altruism as a system of contingent cooperation enforced through reputation and the threat of future non-cooperation. The logic is mechanical: I help you because I predict you'll help me next time.
Psychology reveals the neural substrate: reputation and reciprocity are encoded in the anterior cingulate cortex (ACC) and ventromedial prefrontal cortex (vmPFC), the same regions that process social pain, reward prediction, and the value of social relationships. When you cooperate with someone, their ACC tracks whether they reciprocate. Defectors register as a social loss — your brain codes non-reciprocation as a penalty, not just a neutral outcome.
This neural encoding makes the reciprocity system viscerally felt, not just rationally calculated. You don't just think "they didn't help me back"; you feel betrayed. That feeling is the enforcement mechanism. It makes future cooperation with that person feel aversive — your brain has downweighted their social value.
Where behavioral-mechanics explains the system's logic, psychology explains the neurobiological tooth that makes it work: reputation enforcement through social pain. The threat that drives reciprocal cooperation isn't conscious deliberation about future interactions. It's the ACC's encoding of betrayal as pain, and the amygdala's threat-response to people who've inflicted that pain before.
The tension reveals: reciprocal cooperation works best in small groups where reputation actually travels because the neural system encoding reputation was selected in ancestral environments where everyone's history was common knowledge. In anonymous contexts where reputation doesn't matter, the system fails — not because the logic collapses, but because the neural encoding that normally enforces it (social pain from betrayal) has nothing to attach to.
The Sharpest Implication
Cooperation isn't moral. It's mechanical. You cooperate with people whose cooperation benefits you more than their defection costs. In anonymous contexts where reputation doesn't matter and encounters are one-shot, defection becomes rational. This is why strangers betray each other, why contracts require enforcement, why we build legal systems. Trust requires repetition, reputation, and the threat of future non-cooperation.