When someone asks you to vouch

A former colleague messages you. She is applying for something, and she would like you to put your name to her. You worked together for eighteen months, four years ago. You remember her as pleasant. You remember almost nothing about her work.

You have about ninety seconds to decide, because a longer pause is itself an answer. And every incentive in the situation points the same way: say yes. Saying yes takes a minute, costs nothing today, and keeps a relationship warm. Saying no costs you an awkward exchange, immediately, with a real person you will see again.

So most people say yes. And then, having said yes, they write something that is technically true and carefully empty.

The reason this feels uncomfortable, even when we do it, is that we half-know what we have actually done. A recommendation is not a kindness extended to the person asking. It is a claim about them, made in your name, addressed to someone who is not in the room — the person who will read it and act on it.

That third party is the one who matters, and the one nobody thinks about. They are being asked to accept your judgement in place of evidence they don’t have. They will hire on it, or promote on it, or hand over a team on it. What you owe them is not enthusiasm. It is accuracy.

Notice the shape of the incentive. The cost of saying yes is invisible and deferred; the cost of saying no is visible and immediate. Whenever a system is built that way, it drifts in one direction, and the drift compounds. Once enough people vouch out of politeness, a recommendation stops carrying information — and the person it stops working for is the candidate who genuinely deserved one. The inflation is paid for by the honest.

Which is the part worth sitting with: the generous yes is not generous. It transfers a cost onto two people who never agreed to bear it, in order to spare you ninety seconds of discomfort.

I have refused to give recommendations for two reasons. Sometimes because I did not know the person well enough to make a claim about them. Sometimes because I knew them well enough to know I shouldn’t.

Neither refusal was comfortable, and I do not think it needs to be. But a recommendation is a real instrument, and instruments only work while they still mean something.

The older tradition on this is blunter than we tend to be. Justice was defined, in Roman law and after it, as the settled will to give each person what is actually theirs — not what is pleasant to give, and not what is easiest to withhold. Applied here, it cuts in both directions: the person asking is owed an honest answer, and the person relying on the answer is owed a true one. Neither of them is owed a comfortable one.

There is a middle path, and it is not a compromise — it is the more exact answer. You are rarely asked a yes-or-no question. You are asked to attest to a whole person, when what you actually possess is a fragment.

So give the fragment, and mark its edges. I can speak to how she handled an audit under pressure; I never saw her manage anyone. That is more useful to the person deciding than a warm paragraph, more honest to the person asking, and it is not a refusal. It is the actual answer to the actual question.

And when the honest fragment is too small to be worth anything — say that. Told early and plainly, it costs the person a day. Discovered later, in a reference call that goes flat, it costs them the role.