Dave's reading notes

cooperation-game-theory

Properties1
tagsconcept-society, cooperation, evolution, game-theory, pinker, sapolsky, society

Cooperation and Game Theory

Summary: How does cooperation emerge among self-interested agents? From Tit for Tat in iterated Prisoner’s Dilemma to gossip among hunter-gatherers to the emergence of punishing gods — the same logic recurs at every scale.

Sources: Sapolsky-Behave.md · Pinker-Enlightenment Now.md · Schneier-Liars and Outliers.md

Source pages: Behave: The Biology of Humans at Our Best and Worst — Quotes

Last updated: 2026-04-26

The Prisoner’s Dilemma

In a single-round game, the rational choice is always to defect. In iterated rounds with a known endpoint, backward induction produces the same result — cooperation unravels.

The solution to repeated, open-ended games was found in Axelrod’s tournaments: Tit for Tat — cooperate first, then copy whatever your opponent did last round.

“Tit for Tat can never win [any individual round]. Every other strategy would always beat Tit for Tat by a small margin. However, other strategies playing against each other can produce catastrophic losses. When everything is summed, Tit for Tat wins. It lost nearly every battle but won the war. Or rather, the peace.” (source: Sapolsky-Behave.md)

Tit for Tat is: nice (cooperates first), retaliatory (punishes defection), forgiving (immediately cooperates again after punishment), and clear (easy to read). When signal errors are possible, Forgiving Tit for Tat (occasionally cooperates even after being defected on) outperforms strict TfT.

Reciprocal altruism

Cooperation requires: social species, repeated interactions, individual recognition. “I’ll scratch your back if you scratch mine — and I’m watching you.”

Nomadic hunter-gatherers developed a specialized form: indirect reciprocity. You share now not to be repaid directly but to invest in a reputation for generosity, which pays off when you are in need.

“An HG’s best investment against future hunger is to put meat in other people’s stomachs now.” (source: Sapolsky-Behave.md)

Norm enforcement at scale

Small groups: gossip enforces norms. Large groups (strangers): third-party punishment emerges. Very large groups: morally concerned deities (Big Gods) appear.

“It is only when societies grow large enough that people regularly encounter strangers that ‘Big Gods’ emerge — deities who are concerned with human morality and punish our transgressions.” (source: Sapolsky-Behave.md)

Hunter-gatherer gods, by contrast, are less likely than chance to care whether you have been naughty.

How cooperation starts

In a world of non-cooperators, it is always suboptimal to be the first altruist. Two mechanisms allow cooperation to bootstrap:

  1. Spatial clustering — cooperators can find each other; the cooperative trait facilitates its own spread
  2. Small initial nuclei — a small cluster of cooperators, shielded from defectors, crystallizes outward

“A nidus of cooperation crystallizes outward through the population.” (source: Sapolsky-Behave.md)

Scale without hierarchy

Graeber and Wengrow provide archaeological evidence that large-scale non-hierarchical cooperation was not just theoretically possible but historically documented: Neolithic mega-sites housing thousands of people show no clear evidence of a ruling class or administrative hierarchy. The standard assumption — that cooperation at scale inevitably requires hierarchy to enforce it — is a modern projection, not a historical law. The puzzle is not how hierarchy arose from cooperation, but how and why societies lost the political flexibility to choose otherwise. See graeber-wengrow-dawn-of-everything.

Violence declining

Pinker’s argument (supported but contested by Sapolsky): violence per capita has declined over centuries, driven by the civilizing effects of state institutions, trade, and cultural change. Sapolsky notes that nomadic HGs rarely war; 90–95% of non-state societies engage in warfare. The pacifying effect of state institutions is real but so is the violence that states concentrate when they fail. See violence-and-the-long-peace.

Coopetition: cooperation with rivals

A distinct and practically important form of cooperation is coopetition — simultaneous cooperation and competition with rivals. More than half of coopetitive relationships fail. The performance benefit when they succeed is mediated by absorptive capacity: the firm’s ability to recognize, assimilate, and exploit knowledge flowing through the cooperative channel. Unlearning (shedding outdated routines) catalyzes absorptive capacity but does not independently drive performance.

The Tit for Tat logic applies but is insufficient: with rivals, the calibration problem is not just maintaining cooperation but deciding how much to share at each step without triggering a learning race. See sadeghi-coopetition and coopetition.

Graph View

Backlinks