Premise
Assume that players capable of Generalized Tit-for-Tat (GTfT) and players incapable of GTfT can be identified and separated in advance.
Q1. Pareto Efficiency of Tit-for-Tat
In the iterated Prisoner's Dilemma — whether a two-player game or an n-player game — when all participants adopt the tit-for-tat strategy, every player's long-run payoff is Pareto-efficient compared to any profile where players choose different strategies. (Payoffs exceed those under mutual defection or mixed-strategy equilibria.)
A. True.
Q2. All-Cooperate vs. All-Tit-for-Tat
Given that Q1 holds: when every player cooperates unconditionally versus when every player plays tit-for-tat, are the expected payoffs identical?
A. Yes. In a population where no one defects, tit-for-tat behaves identically to unconditional cooperation — every round produces mutual cooperation. The expected payoffs coincide. However, tit-for-tat is strictly more robust against the entry of external defectors: a defector exploiting an unconditional cooperator earns the temptation payoff indefinitely, whereas against tit-for-tat the exploitation is limited to a single round.
Q3. Optimal Response to Non-GTfT Players
When some players cannot play GTfT, do GTfT-capable players maximize their payoffs by playing tit-for-tat exclusively with other GTfT-capable players and always defecting against — or refusing to transact with — non-GTfT players?
A. Yes. The payoff of GTfT-capable players is maximized under this policy. Cooperating with a player who cannot reciprocate tit-for-tat exposes the cooperator to exploitation without the guarantee of future retaliation. Defecting against (or excluding) such players eliminates this downside risk while preserving the cooperative surplus within the GTfT-capable group.
Two-Layer Network Structure
If Q1–Q3 hold, the payoff-maximizing network for GTfT-capable players is a two-layer structure:
| | Layer | Interaction | Payoff character | |---|---|---|---| | Layer 1 | GTfT-capable ↔ GTfT-capable | GTfT (mutual cooperation equilibrium) | High payoff | | Layer 2 | GTfT-capable ↔ GTfT-incapable | AllD (exploitation or loss minimization) | Loss-minimized |
Why This Structure is Optimal
-
Layer 1 sustains the cooperative equilibrium. Because every participant can and does play tit-for-tat, defection is immediately punished and cooperation is self-enforcing. The resulting payoff profile is Pareto-efficient (Q1) and as high as unconditional cooperation but strictly more robust (Q2).
-
Layer 2 eliminates the vulnerability. Non-GTfT players — whether they always defect, cooperate randomly, or follow some other non-reciprocal strategy — cannot sustain the punishment mechanism that makes cooperation incentive-compatible. Playing AllD against them (or refusing interaction entirely) ensures that GTfT-capable players never subsidize non-reciprocal behavior (Q3).
-
The boundary between layers is maintained by the premise: GTfT capability is observable ex ante. This identification step is what makes the entire structure feasible. Without it, cooperators cannot distinguish reciprocators from exploiters, and the cooperative equilibrium unravels.
Implication
The generalized optimal network is not one where everyone cooperates. It is one where cooperation is conditional and bounded: extended fully within a verified reciprocal group and withheld entirely outside it. The network's optimality derives not from universal goodwill but from the structural enforcement of reciprocity.