Strategic decisions are rarely one-shot. They repeat, they unfold over a project's life, and the other parties learn and adapt just as you do. That single change — a future — transforms what's rational.
Extend the game of Chapter 23 across time and you get a Markov game (a stochastic game): a shared situation that evolves based on the joint actions of all agents, with each agent collecting its own rewards over the sequence. It is the MDP of Part II with more than one decision-maker — and the added twist that everyone is adapting simultaneously.
1The shadow of the future
The most important insight here is why repetition changes everything. In a one-off prisoner's dilemma, defection is rational. But when the interaction repeats — as nearly every real project relationship does — cooperation can become the rational choice, because defecting today invites retaliation tomorrow. This is the shadow of the future: the value of the ongoing relationship disciplines present behaviour. A well-known result confirms that a wide range of cooperative outcomes can be sustained in repeated play, held in place by the credible threat of future response.
2Learning against a moving target
When agents don't know the game and must learn — multiagent reinforcement learning — a new difficulty appears. Each agent is learning while the others learn too, so from any one agent's view the environment keeps changing: the very thing it's adapting to is itself adapting. This non-stationarity is what makes multiagent learning genuinely hard, and it's why naïvely applying single-agent methods can chase its own tail. Techniques that anticipate others' adaptation, or that converge to equilibrium play, are needed to make progress.
A future changes the game. Cooperation that's irrational in a one-shot encounter becomes rational when the relationship continues and defection can be punished. And when everyone is adapting at once, you're optimising against a moving target — the core challenge of learning among others.
The single most useful lever in multi-party project work is lengthening the shadow of the future: make relationships ongoing, outcomes repeated, and reputations visible, so that cooperating pays and defecting costs. This is precisely why long-term framework partnerships tend to behave better than one-off, lowest-price contracts — the future keeps everyone honest. And when you're negotiating against a party that keeps shifting its approach, don't expect a fixed opponent; anticipate that they're adapting to you, and plan for a moving target.