Strategic Interaction and Cooperation
On this page
Prerequisites: Economic Choice and Analysis · Surplus, Taxes, and Subsidies
Two teams share a single interface. Both parties maintain the documentation diligently, leading to the smoothest long-term collaboration; however, if the other party has already written the documentation, saving maintenance time and directly using their output may yield higher immediate benefits. The question is no longer “what does my action alone produce,” but rather “given that the other party is also making a choice, what is my best action?”
This article first asks you to make a choice in a single interaction, then checks the incentives column by column, and finally alters the future opportunities for cooperation and observability. The following payoffs are fictional net benefit points, which have already integrated the current investment and usage value. The numbers serve only for comparison and do not represent the teams’ actual performance.
Choose first, then examine the incentives of both parties
Each team can choose to Maintain (C) or Save Effort (D). In the matrix, the first number is Team A’s payoff, and the second is Team B’s payoff. Both choose simultaneously, and no enforceable maintenance contract can be signed beforehand.
| Team A's Choice / Team B's Choice | B Maintains C | B Saves Effort D |
|---|---|---|
| A Maintains C | 3, 3 | 0, 5 |
| A Saves Effort D | 5, 0 | 1, 1 |
Stop for a moment: If you were Team A, how would you choose? The reasoning must explicitly address one of Team B’s choices, not just say “cooperation is best for everyone.”
Fixing Team B’s choice to Maintain, Team A compares the same column: Maintaining yields 3, Saving Effort yields 5, so Saving Effort is better. Fixing Team B’s choice to Save Effort, Team A compares the other column: Maintaining yields 0, Saving Effort yields 1, so Saving Effort is still better. In this table, D is a strictly dominant strategy for Team A, outperforming C regardless of what Team B does. Team B faces a symmetric structure and will compare in the same way.
Thus, both parties choose to Save Effort, resulting in payoffs of 1, 1. At this position, if Team A unilaterally switches to Maintaining, the payoff drops from 1 to 0, and the same applies to Team B; no one can increase their payoff by unilaterally changing their action, so this is a Nash equilibrium.
A Nash equilibrium does not imply maximum payoff, moral superiority, or that reality will necessarily reach this state immediately. It only checks whether one would still want to change given the other party’s action. If both parties jointly switched to Maintaining, they would achieve 3, 3, but “getting better together” does not mean “each person benefits from changing alone.”
Why verbal agreements may be insufficient
The two teams meet and agree to maintain the documentation, but this does not automatically change the matrix. When it comes to execution, if Team B maintains as agreed, Team A can still get 5 by Saving Effort instead of 3. Without future consequences, accountability constraints, or different preferences, pre-commitment cannot eliminate the incentive to deviate ex post.
This does not mean people are always selfish or that cooperation is impossible. It shows that the current model omits certain payoffs: reputation, professional norms, costs of breach, and concern for colleagues may all influence actual choices. To explain cooperation, one must explicitly include relevant factors rather than using an old payoff table while demanding that people naturally choose another outcome.
For example, independent verification can detect unmaintained documentation and deduct 3 payoff points for Saving Effort: When Team B Maintains, Team A gets only 2 by Saving Effort, which is lower than the 3 from Maintaining; when Team B Saves Effort, Team A gets −2 by Saving Effort, which is lower than the 0 from Maintaining. In the new table, Maintaining becomes the dominant choice. Whether this rule is worth adopting requires deducting verification costs and checking whether behavior can be reliably identified; one cannot treat punishment as a cost-free switch.
Will the same people cooperate in the future?
Now change one condition: The two teams must collaborate every period, and current behavior is accurately observed. Consider a conditional cooperation strategy: If both parties have maintained previously, continue to maintain; if anyone Saves Effort, both will Save Effort thereafter. Let the weight of future period payoffs relative to the present be δ, with 0≤δ<1.
The payoff of continuous maintenance is 3+3δ+3δ²+…=3/(1−δ). If one secretly Saves Effort, they get 5 once, and then 1 per period thereafter, with a payoff of 5+δ/(1−δ). For Maintaining to be at least as good as deviating, we need:
3/(1−δ) ≥ 5+δ/(1−δ)
3 ≥ 5−4δ
δ ≥ 0.5
For example, if δ=0.8, the value of continuous maintenance is 15, while deviation yields 9; when future relationships are sufficiently important, gaining 2 extra points now is not worth it. If δ=0.2, continuous maintenance yields 3.75, while deviation yields 5.25; if the punishment is too distant or the relationship too unstable, it is insufficient to support cooperation.
This is merely a conditional test given specific strategies, accurate observation, and an infinite or uncertain horizon; it is not a universal threshold for all team cooperation. Repeated interaction does not guarantee that cooperation will definitely emerge; both parties may also fail to coordinate on the same strategy.
Putting false positives and relationship endpoints back in
If the documentation platform occasionally loses submissions, a single technical glitch might be mistaken for Saving Effort, and permanent punishment would destroy cooperation that could otherwise continue. Now, besides preventing deviation, there is a need for appeals, resubmissions, evidence verification, or limited punishment. Tolerance reduces collateral damage but may also be exploited; rules must balance between these two issues.
If both parties know with certainty that there is only one final period remaining, and the original pure payoff structure is maintained, there is no future punishment in the last period, so they will still Save Effort; backward induction affects choices in earlier periods. This finite-horizon conclusion relies on conditions such as a clear endpoint, common knowledge, and perfect rationality. Real organizations have subsequent reputations, cross-project relationships, and incomplete information, so one cannot simply conclude that cooperation will definitely not occur just because “the project is ending.”
Sequential choice requires drawing a decision tree separately. If Team A invests first, and Team B decides after observing, Team A must anticipate Team B’s best response at that time; one cannot treat “I hope he reciprocates” as a credible commitment. Contracts, deposits, and milestone verifications serve precisely to change the feasible choices and payoffs at certain future nodes.
Separate equilibrium from evaluation
This table contains only the net payoffs of the two teams. Suppose their cooperation harms a third party, for example, by jointly restricting services and causing users to pay higher prices; then an increase in the two parties’ payoffs does not mean social welfare improves. Game analysis first explains participant behavior; evaluation requires adding the excluded subjects.
Similarly, an equilibrium may rely on incorrect information, high switching costs, or unfair initial rights. Finding an equilibrium is merely an analytical step and cannot replace discussion of the rules themselves. The next lesson on information and contracts will continue to handle transactions when “what type the other party is” and “what the other party did later” are unobservable.
Answer using comparison rather than labels
Change the payoff for both parties Maintaining from 3, 3 to 6, 6, leaving the other three cells unchanged. Is Saving Effort still Team A’s dominant strategy? Is both parties Saving Effort still a Nash equilibrium?
Expand reasoning
Fixing Team B’s choice to Maintain, Team A gets 6 by Maintaining and 5 by Saving Effort, so Team A should Maintain; fixing Team B’s choice to Save Effort, Team A should still Save Effort. Therefore, Saving Effort is no longer dominant. Both parties Saving Effort remains a Nash equilibrium, because unilateral switching to Maintaining would drop the payoff from 1 to 0; both parties Maintaining also becomes a Nash equilibrium, because deviation would drop the payoff from 6 to 5. At this point, a coordination problem arises, and one can no longer simply apply the original unique dominant result.
The manager says, “We will no longer work with teams that Save Effort,” but that team holds the only equipment capable of fixing the fault. Is this threat necessarily credible? How do you need to supplement the model?
Expand reasoning
Not necessarily. To assess credibility, one must compare the manager’s choices at the future node where repair is truly needed: which gives the manager the higher net payoff: refusing to cooperate or continuing cooperation, whether alternative equipment exists, and whether the commitment has enforceable constraints. If the manager always chooses to continue cooperating when the time comes, the other party will anticipate this. Credibility must be supported by incentives at future nodes, not by how harsh the wording is.
Sources and Further Reading
These references support concepts and statistical definitions; the numerical cases and diagrams are original synthetic teaching examples.