Four Strategies for Trust, Compared
Trust has been studied experimentally for decades, largely by giving people money and seeing what they do with it. The results are more encouraging than most people expect, and they point at a specific set of behaviours rather than a disposition.
Four strategies for handling trust, compared on what the evidence says each one actually produces.
Strategy one: send nothing
The standard experiment works like this. You are given a sum and may send any portion to a stranger. Whatever you send is tripled. The stranger may then return any portion, or none.
Pure self-interest predicts sending nothing, because a rational stranger will return nothing and you would lose everything you sent.
What actually happens is that most people send around half, and most recipients return enough that senders come out ahead on average. The prediction of zero is wrong, consistently, across many replications and many countries.
What this strategy costs: the tripling. Sending nothing guarantees you keep what you have and forfeit the entire gain that only exists because the money moved. In a one-shot game it is safe; in any repeated setting it is the strategy that never discovers anything.
Strategy two: always cooperate
The opposite extreme, and the one usually described as trusting.
Against cooperative partners it does well. Against exploitative ones it does badly and keeps doing badly, because nothing about it changes in response to being taken advantage of.
What this strategy costs: it removes any incentive for the other party to behave well, which over time selects for the partners least worth trusting. Unconditional trust is not generous so much as uninformative — it produces no signal either party can learn from.
Strategy three: retaliate and hold it
Cooperate until defected against, then defect permanently.
This protects you, and in tournaments where strategies played each other repeatedly it performed poorly. The reason is that it cannot recover. A single defection — including an accidental one, or a misread — locks two parties into mutual punishment for the remainder of the relationship.
What this strategy costs: everything after the first mistake. In any long relationship, misunderstandings are certain, and a strategy with no path back converts an error into a permanent state.
Strategy four: cooperate, retaliate once, forgive
The strategy that won Axelrod's computer tournaments was also the simplest submitted: open by cooperating, then do whatever the other party did last time.
Its four properties have held up as a description of what works:
Nice — it never defects first. Across the tournaments, the strategies that opened cooperatively outperformed those that did not.
Retaliatory — it responds immediately to defection, which prevents exploitation.
Forgiving — one act of cooperation from the other side restores the relationship completely. It does not carry history.
Clear — the other party can work out the rule quickly, which lets them adapt. Strategies that were complex or unpredictable did worse, because nobody could learn how to succeed with them.
What this strategy costs: it gets defected against once by anyone who tries. That is the price of finding out, and it is the lowest price on this list.
The property people find hardest
Of the four, forgiveness is the one that gets dropped in practice, and it is the one doing the most work.
A grudge feels like protection. In the tournament results it functioned as a liability, because it removed the possibility of returning to mutual benefit after a single lapse. Retaliation and grudge-holding are different behaviours: the first is a response, the second is a state.
Clarity is a close second. People frequently withdraw trust without indicating that anything has changed, which leaves the other party unable to correct. An unexplained withdrawal is exactly the unpredictable strategy that performed worst.
What the experiments do not settle
Two honest limits.
These are games with defined payoffs and symmetric power. Real relationships have neither, and where power is asymmetric the calculus changes substantially — the party who can absorb a defection is not in the same position as the party who cannot.
And tournament results describe what performs well across many interactions. They do not tell you whether to trust a specific person once, which is the question people usually have.
The thirty minutes
Make a cup and take one relationship — professional or personal — through the four strategies. Which one are you actually running?
Most people discover they are running the third: cooperative until something went wrong, then quietly permanent about it. It rarely feels like a strategy from the inside. It feels like an accurate assessment.
Then three questions. Was there one defection or a pattern — the difference determines whether forgiveness is warranted or naive. Does the other party know your position has changed, or are you running an unreadable strategy. And what would opening cooperatively cost you now, concretely, given that the evidence says the cost is usually one exchange rather than everything.
Half an hour is enough for one relationship. It is not enough for a general policy on trust, and anyone who arrives at one in a single sitting has probably reasoned from a bad recent experience.
An aside on clarity, since it was one of the four properties: it applies to what you serve people too. Our blends differ enough that saying which one it is beats letting someone guess.
