The state of nature

Why would people who all sincerely want peace still rationally arm themselves against each other?

The usual explanation for conflict is that somebody wanted it. There was an aggressor, or a fanatic, or a fool. Remove them and the trouble goes away.

Hobbes's claim in Leviathan (1651) is stranger and harder to shake off: that you can populate a world entirely with people who sincerely prefer peace, give none of them any appetite for conquest, and still get a general condition of armed fear. If that is right, then the conflict is not coming out of anyone's wishes. It is coming out of the arrangement they are all standing in. So the question is what feature of an arrangement could manufacture an outcome that every single participant would vote against.

Begin with one person, deciding alone. You want peace and you are unarmed. What do you actually need to know before that is a safe position? Not what your neighbour wants — what he might do. And you cannot see wants; you can only see capacities and hear claims.

Suppose he tells you, sincerely, that he means you no harm. Is that enough? Ask what would have to hold for it to be enough. He would have to be telling the truth, and to still mean it next year, and to be strong enough not to be forced into it by a third party. Notice that even a completely honest promise fails the third test. So the assurance you need is not about his character at all.

Now hold that thought and let everyone reason the same way at once — because they will, and each of them is doing something perfectly reasonable. You arm, defensively. He sees the arming, cannot see the "defensively", and arms in response. Each step is prudent, each step is honest, and the sequence ends with everyone armed and everyone more frightened than when they began. Hobbes called this middle cause diffidence — anticipation — and he put it alongside competition and glory precisely because it needs neither greed nor pride to run. It runs on nothing but the absence of assurance.

This is worth naming as the type of problem it is. It is not a case where each person's selfish interest beats the common one; if it were, offering people a better deal would fix it. Every person's first choice here is mutual restraint. What fails is not preference but expectation. Rousseau's stag hunt, in the 1755 Discourse on Inequality, makes the shape visible: a group can take the stag only if all of them hold their positions, and any one of them who doubts the others will do better to grab the hare in front of him. His doubt, not his appetite, is what loses the stag.

So what exactly is the good that cannot be produced? Not peace — peace is easy if everyone holds. The good that no one can produce is the belief that everyone will hold. And that good has the awkward property that each person's contribution to it is worthless on its own. My restraint does not make you confident, because you know I would abandon it the moment I feared you. Nobody defects here out of malice; the good simply cannot be assembled from individual pieces.

Which points at what would have to change. If the missing ingredient is assurance rather than goodwill, then more goodwill cannot supply it, and neither can more promises — Hobbes's line is that covenants without the sword are but words. What is needed is something that makes defection unprofitable for everyone simultaneously, so that each person can predict the others without trusting them. That is the whole argument for a common power, and it explains a feature of the argument that often looks like a mistake: the sovereign is not there to make people good. It is there to make them predictable.

Two honest caveats. First, Hobbes did not mainly claim this as a lost historical era; he said the standing example is sovereigns themselves, who stand toward each other in the posture of gladiators, and the modern international relations literature analyses that case under the name security dilemma. Second, the inference from the argument to absolute sovereignty is very much disputed — Locke thought a much lighter authority sufficed, and later writers have documented communities that solve assurance problems locally, without a sovereign, through repeated dealing and reputation. The diagnosis has worn far better than the prescription.

The most common misreading is that Hobbes thought people were nasty. He did not need that, and the argument is stronger without it — its force comes from showing that decent, peace-loving people generate the same outcome. A second misreading is that arming is irrational because everyone ends up worse off. Each individual decision is entirely rational given what that person can know; the irrationality lives in the aggregate, which is exactly what makes it a collective-action problem rather than a mistake anyone could be talked out of. Third, the exit is not everyone deciding to trust one another. It is everyone becoming subject to something that makes trust unnecessary.

• The stag hunt and the assurance game against the prisoner's dilemma: why the difference changes which remedies can work.
• Locke's much milder state of nature, and what he thought was already available in it.
• Elinor Ostrom's cases of communities solving assurance problems without a sovereign.
• The security dilemma as the same structure between states, where no sovereign is on offer.

Every term the collection defines is gathered in the glossary.