Two Organizations Can Be Allies on Monday and Enemies on Tuesday, and Both Are Sincere

We describe relations between organisations the way we describe relations between people. Two countries are friends or rivals. Two companies are partners or competitors. Two ministries are on the same side or not. The vocabulary comes from a world of individuals, where a friend is a friend across the board, and it is wrong for organisations in a way that produces bad predictions daily.

An organisation has interests, not friendships, and interests are indexed by issue. Two organisations agree wherever their interests coincide and fight wherever they diverge, and there is nothing inconsistent about doing both in the same week.

The Pattern

The United States and China trade on a scale no two countries ever have, and each is the other's principal military concern. Neither fact is a mask for the other. On the question of whether goods should flow, the interests align; on the question of who controls the western Pacific, they do not. Commentators who insist on knowing whether the two are really friends or really enemies are asking a question that has no answer, because the entity they are asking about does not have a single disposition.

Closer to home, a finance ministry and a defence ministry cooperate completely on getting the budget passed and then fight to the last shekel over how it is divided. A hospital and an insurer share an interest in the patient paying and oppose each other on who receives the money. Two universities co-author a grant application in the morning and poach each other's faculty in the afternoon.

None of these actors is being two-faced. Each is behaving exactly as it should given what it is.

Why Individuals Are Different

A person has a small number of relationships and a reputation that spans all of them. If I cooperate with you on one matter and undermine you on another, you experience it as betrayal, and you are right to, because for individuals the relationship is the unit. I have argued before that this equipment evolved in small groups where the same few people met over everything, and in that setting an across-the-board loyalty is the rational strategy.

An organisation is not in that setting. It meets hundreds of counterparts across hundreds of questions, and its decisions are made by different people in different departments who may never speak. The trade negotiator and the admiral are both acting for the same state, and neither is instructed to take the other's game into account. The organisation has no place where the friendship could be stored even if it wanted one.

So the game-theoretic description is not one repeated game with one counterpart. It is a portfolio of separate games, each with its own payoff matrix, running in parallel, and the only thing they share is a name.

What This Predicts

Several things that the friendship model gets wrong.

Alliances will look unstable and will not be. A state that votes with you at the United Nations and against you at the trade organisation is not wavering. It is behaving consistently on each issue, and the appearance of inconsistency is an artefact of your expecting one disposition.

Appeals to loyalty will fail. Reminding a counterpart organisation of past cooperation on one matter buys nothing on another, because the past cooperation was not a favour; it was the intersection of interests on that matter, and the intersection on this one is empty.

Personal relationships between the heads of organisations will matter less than everyone hopes. Two leaders can genuinely like each other and their organisations will still fight wherever the payoffs say fight, because the leaders are not the ones making most of the decisions and cannot override the ones they do not make.

And the way to move an organisation is not to become its friend. It is to change the payoff on the specific question you care about, so that its interest on that question moves. This is less pleasant than diplomacy and much more effective.

The One Place the Model Breaks

There is a case where organisations do behave as if they had a single disposition, and it is worth naming because it is the exception that people mistake for the rule.

When one issue becomes existential, it swallows the others. A state at war for its survival subordinates trade, culture and everything else to the one game that matters, and for the duration it does have a single disposition toward its enemy. The friendship model is a description of organisations in that state, generalised to organisations in every other state, where it does not apply.

Most of the time no issue is existential, the portfolio is a portfolio, and the honest question about any two organisations is never whether they are allies. It is: on which issues, this year, do their interests coincide, and on which do they not? That question has an answer, and the answer changes, and changing is not betrayal. It is what an organisation is.