The strongest version of the argument

There is a familiar piece of advice about disagreement: before you criticise a position, state it in the form its most capable defender would recognise. The practice usually goes by steelmanning, and it is usually justified on grounds of fairness. You owe your opponent an honest hearing. Attacking a caricature is a species of bad manners.

That justification is true and, I think, almost beside the point. The reason to steelman is not that you owe it to the other side. It is that you owe it to yourself, because the strawman version tells you nothing about whether you are right.

What refutation is for

Consider what you are actually doing when you argue against a position. You are running a test. The hypothesis is my view is correct and theirs is not, and the test consists of confronting your view with the best available reason to abandon it.

A test is only informative if it could have come out the other way. If you construct an opponent too weak to threaten your position, the outcome was fixed before you began, and passing tells you exactly nothing you didn't already believe. You have not learned that you are right. You have learned that you are capable of inventing someone who is wrong.

A refutation that could not have failed is not a refutation. It is a ceremony.

This is why the practice cannot be reduced to politeness. Politeness is about how the exchange feels to the other person. Steelmanning is about whether the exchange contains any information at all.

Three ways it goes wrong

The advice is easy to state and hard to follow, and it fails in recognisable ways.

Substituting the weakest actual holder for the strongest available argument. It is always possible to find someone defending your opponent's conclusion for terrible reasons. Quoting them is not engagement; the question is whether the position survives, not whether its worst advocate does.

Reconstructing the argument in your own vocabulary. Positions are often carried by their terminology. Translate a claim into the terms of the framework you are about to judge it by, and you have frequently assumed the conclusion before the argument starts. The test is whether a thoughtful adherent would recognise your restatement — not whether it sounds reasonable to you.

Steelmanning the conclusion instead of the reasoning. This is the subtlest failure. You find a good argument for their conclusion — one they never made — refute that, and declare the matter closed. But the reasoning people actually use is what determines what else they believe and what would change their minds. Defeating a substitute argument leaves the real one untouched.

The cost, which is real

Done properly, this is expensive. It requires reading people you find tiresome, and it requires holding a position in mind sympathetically for long enough to see what makes it attractive — which is uncomfortable, because sympathetic attention is how minds get changed.

That discomfort is not a side effect. It is the mechanism. A procedure that could never move you is not a procedure for finding out anything.

There is also a limit worth naming. Not every position has a strong version, and the demand to find one can become its own bad-faith move — an infinite obligation to keep improving your opponent's case until you have manufactured a view no one holds. The honest form of the practice includes a stopping point: here is the best version I can construct; here is why it fails; if there is a better one, I have not found it.

That last clause is the whole thing. It is what separates an argument from an announcement.

Why it matters more in politics

In most domains, believing false things has a cost that eventually finds you. Political beliefs are unusual in that almost none of them are ever tested against reality by the person holding them. Your vote does not decide the election. Your opinion on a policy will never be checked against what that policy does.

What that means is that the ordinary corrective — being wrong and finding out — is largely unavailable. The only correction left is argument. And argument only corrects anything if it is conducted against the real thing.

Which is to say: the reason to take the other side seriously is not that they deserve it. Sometimes they don't. It is that a habit of arguing against the weakest available opponent will, over years, leave you confidently holding a set of views you have never actually examined — and no way to tell the difference between that and knowledge.