Summary

Axelrod reduces cooperation to an iterated prisoner’s dilemma: when the future is important, egoists can learn to cooperate without central authority, friendship, trust, threats, or enforceable commitments. The durable strategy is tit for tat: nice, provocable, forgiving, and clear. Cooperation emerges when relationships repeat, payoffs reward reciprocity, and players can recognize patterns.

“The foundation of cooperation is not really trust, but the durability of the relationship”

Key Takeaways

Notes

Introduction

• Inspiration: understanding iterated prisoners dilemma (ran original tit-for-tat computer program tournament)

1: Introduction

1: The Problem of Cooperation

• “Under which condition will cooperation emerge among a world of egoists with no central authority?”
• Hobbes famously argues a centralized gov’t is necessary
• ↑ though, doesn’t apply to nations or really anything without laws (eg. social norms / custom)
• Whole book will revolve around the prisoners dilemma
• In finite games of known length without comms, best move is always defect
• In indefinite length games, cooperation becomes best
• Thus, what are the necessary and precise conditions?
• Simplifications:
  • No threats or enforceable commitments
  • No way to be sure what the other will do (meta-game analysis)
  • No reputation
  • No way to eliminate other player or run away from interaction
  • No way to change the payoffs
• w = discount parameter for future moves
• Unlike chess, players are not entirely antagonistic
• If future is important (w is high), there is no best strategy independent of the best strategy used by the other player
• 4 properties generally advantageous:
  • Avoidance of unnecessary conflict
  • Provocability in the face of uncalled for defection by other player
  • Forgiveness after responding to a provocation
  • Clarity of behavior so the other player can adapt to your pattern
• Evolution and survival of the fittest can teach cooperation
• 4 simple suggestions:
  • Do not be envious of others’ success
  • Don’t be the first to defect
  • Reciprocate both cooperation and defection
  • Don’t be too clever
• Gov’t must make it profitable to cooperate most of the time
• Evolution of cooperation speeds up if participants know the rules of the game (laws)

2: The Emergence of Cooperation

2: The Success of “tit for tat”

• “Nice”: never defects first
• Key insight: minimize echo effects of mutual defection
• Even better: tit for tit for tat (be more forgiving)
• Using GAs / evolution of rules, can prove you need to do well against “nicer” rules - else, poison your own environment

3: The Chronology of Cooperation

• Tit for tat is collectively stable if and only if w param is large enough
  • If w < 2/3, pays to defect on alt moves
  • If w < 1/2, pays to defect on every move
• Any strategy which may be first to cooperate is collectively stable only when w is sufficiently large
• For a “nice” strategy to be collectively stable, it must be provoked on the very first defection
• Always defect is collectively stable, though can be invaded by pockets of “nice” strategies

3: Cooperation Without Friendship or Foresight

4: WW1 Trench Warfare: Live-and-Let-Live

• Zero-sum iterated prisoners dilemma
• Retaliation, patterns of behavior
• HQ solution for oversight: raids

5: Evolution of Cooperation in Biological Systems

• Genetical kinship theory: genes-eye view
• Reciprocity theory:
• Kinship can lead to reciprocity evolving
• Mutualism*
• Eg. gut bacteria, cancer dormant can trigger when host is weak

4: Advice for Participants and Reformers

6: How to Choose Effectively

• 4 simple suggestions:
  • Do not be envious of others’ success
  • Don’t be the first to defect
  • Reciprocate both cooperation and defection
  • Don’t be too clever
• Tit for tat never wins a single game, but is good at eliciting cooperative behavior in others
• Don’t over-game theory - consider reverberation
• Don’t make enemies
• Make it clear you will reciprocate

7: How to Promote Cooperation

• 3 categories to promote cooperation:
  • Make the future more important relative to the present
  • Change the payoffs for the 4 possible outcomes of a move
  • Teaching the players values, facts, and skills that promote cooperation
• Make interactions more durable and more frequent
  • Eg. decompose a big agreement into smaller steps
• Gov’ts + laws change payoffs
• Free rider problem
• Improve recognition abilities (transparency)

5: Conclusions

8: The Social Structure of Cooperation

• 4 factors which can give rise to interesting social structures:
  • Labels: fixed characteristics observable by other players before interaction begins
    • Stereotypes
    • Status hierarchies
  • Reputation: best is being a bully / credible threat
    • Eg. tax collection, monopolies
  • Regulation
    • Eg. child support for visitation
  • Territoriality
    • Geographic or abstract (eg. similar products, ideology)
    • Think of as a territorial 2D grid
    • Can introduce local collectively stable strategies

9: The Robustness of Reciprocity

• “The foundation of cooperation is not really trust, but the durability of the relationship”
• Higher turnover = lessening chance of cooperation, even within an org
• Ratchet effect: reciprocity norm is self-reinforcing

Afterword

• Shortcoming: other player misunderstanding your choice
• Better solution: generous tit-for-tat (10% chance of not reciprocally defecting) would’ve won the tournament
• You can win the tournament without winning a single game
• Norms are as important as laws
• Methods for mass cooperation:
  • Share your own experiences → incentive to develop a reputation for cooperation
  • Refuse to interact with non-cooperators
  • Social capital: build networks of trust and reciprocity (Bowling Alone)