A client asks for a deposit on work you already suspect you will not be able to deliver. Taking it solves this month. You tell yourself it is an exception.

Kant's test is the one that catches this, and it catches it without appealing to consequences, to your character, or to how it would feel if you were found out.

Two kinds of ought

Almost every rule you follow is conditional. If you want the client to stay, answer quickly. If you want to be trusted, do not lie. These hold only while you want the thing. Stop wanting it and the rule releases you.

Kant called these hypothetical imperatives, and his claim is that genuine moral demands are not like this. Do not take money for work you cannot do is not advice about how to get what you want. It holds whether or not it serves you. If morality bottomed out in what you happen to want, it would not be morality — it would be strategy with a better reputation.

So: is there a demand that binds you simply because you are rational, with no if in front of it? Kant said yes, one, and everything else follows from its form.

The test, and the two ways it fails

Act only on a maxim you could will to be a universal law.

Not would you like it if everyone did this — that is a question about your preferences. The question is whether the universalised version is coherent. And Kant identified two distinct ways it breaks.

It destroys its own conditions. Universalise the deposit. If everyone took payment for work they could not deliver whenever it was convenient, deposits would stop being given — nobody would extend the trust. The maxim depends on an institution that the maxim, generalized, dissolves. You cannot will it universally because there would be nothing left to exploit. Kant calls this a contradiction in conception, and it is the sharper of the two.

You cannot coherently want it. Universalise letting your abilities decay. Nothing collapses; a world of people who never developed anything is possible. But you cannot consistently will both that you are a being with purposes and that you systematically destroy your capacity to pursue any. This is a contradiction in the will, and it is weaker — critics have pushed on it for two centuries.

The formulation people actually use

Kant gave several versions and insisted they were the same law seen from different sides. One of them escaped philosophy entirely:

Act so that you treat humanity, in yourself and in others, always also as an end, and never merely as a means.

Read the merely carefully — it does the work. You use people as means constantly and there is nothing wrong with it. The person who fixed your car was a means to a working car. What the formula forbids is treating someone as only that — as a resource with no purposes of their own that you are required to take seriously.

This is the sentence underneath the modern idea of human dignity. It is what people are reaching for when they say someone was treated like a number, and it is the test that makes a layoff handled honestly different from the same layoff handled as inventory.

The objection everyone arrives with

A murderer comes to your door and asks where your friend is hiding. Kant addressed this in 1797 and said you may not lie.

This has done more damage to his reputation than any criticism by an opponent, and it is worth being precise about what went wrong. The problem is Kant's own application of the test — specifically the decision to describe the maxim as lying rather than as deceiving someone who intends murder to do it.

Universalise the second, properly specified, and nothing collapses. The institution of truthfulness is not threatened by a rule that people may mislead those attempting homicide. Contemporary Kantians — Korsgaard and O'Neill among them — argue the framework survives the case and that Kant simply misapplied his own instrument.

Whether specifying maxims that finely leaves the test with any teeth is the live question. Do it too freely and every maxim can be narrowed until it universalises.

Why "from duty" and not "from feeling"

The most alien part of Kant, and the part where the alienness is the point.

An action has moral worth, he argues, when the will is determined by recognition of the law — not by inclination, not by command, not by anticipated payoff. He calls this autonomy: a will that legislates to itself. A will moved by anything else is heteronomous, however pleasant the thing moving it.

The notorious consequence is that helping someone because you feel for them has no moral worth, while helping them because you recognize you must does. Schiller mocked it and people have been uncomfortable ever since.

The defensible version: sympathy is not reliable. It tracks proximity, attractiveness, and resemblance to yourself. A person who is kind only when moved has outsourced their conduct to their moods. Kant is not saying feeling is bad. He is saying it cannot be what the obligation rests on, because it will not be there when it is needed most.

Why it is worth your time

Most ethical reasoning in practice is consequentialist — weigh the outcomes, pick the best. It works until you are the one doing the weighing and you benefit from one of the answers.

The universalisation test is useful precisely because it does not ask about outcomes and cannot be bent by your stake in them. It asks a structural question: what is the rule I am actually operating on, and does it survive being made general?

Run it on any arrangement that works for you only because other people are not doing it. That is the whole diagnostic, and it takes about ten seconds.

Further reading