Steel-manning is the deliberate practice of stating an opponent's position in the strongest, most defensible form they themselves might recognize and endorse, before offering any refutation. The name is a direct inversion of straw-manning, the fallacy of attacking a weakened distortion. Where the straw man wins by misrepresenting, the steel man wins by getting the argument right and beating it anyway. The first audience always knows when you've cheated. The steel-man removes the suspicion that you're cheating.
If you can pass an ideological Turing test for the other side, your refutation costs nothing in credibility. If you cannot, every word of your refutation is downgraded by audiences who know the position better than you do.
How It Works
Most arguments fail not because the opposition is unbeatable but because the rebuttal targets a position the opponent never held. The audience hears the original argument; they hear the rebuttal; they notice the gap; the rebuttal is dismissed. Steel-manning closes that gap. By stating the opposing view at least as well as its strongest defender would state it, you signal three things to the audience at once: you understand the position, you respect the people who hold it, and you have engaged with the actual argument rather than a convenient version of it. Only after that signal is established does the rebuttal land.
The technique is older than its name. Aristotle counsels in the Topics that the dialectician should know the strongest objections to their own thesis before defending it. John Stuart Mill, in On Liberty, argues that no one understands their own position fully until they have heard it disputed by its ablest opponents. The principle of charity in twentieth-century philosophy formalized the requirement: in interpreting any text or speaker, attribute the most coherent and true position consistent with their words. Steel-manning is principle-of-charity weaponized for adversarial debate.
How to Build a Steel Man
A steel man has three components and they are not optional. Skip any one and the structure collapses back into a straw man dressed up.
- State the position in its proponent's own framing. Use their vocabulary, their distinctions, their canonical examples. If they say "earned benefits" you do not say "entitlement spending." If they say "common-sense regulation" you do not say "government overreach." Renaming the territory is the first move of straw-manning.
- Add the strongest evidence the position has. Not "some studies suggest" but the specific, named studies the proponents cite. Not a vague gesture at history but the historical case that most directly supports their claim. The audience must believe you have done the reading.
- Acknowledge the appeal. Most defensible positions have a moral or intuitive pull beyond the empirical case. Name it. "The appeal of this view is that..." Do not skip to refutation while the audience is still feeling the appeal.
Only after these three are in place do you transition to the rebuttal. Common transition phrases: "Even taking this position at its strongest, however...", "Granting all of this, the difficulty is...", "The position has more force than its critics usually acknowledge, but it still fails because..." The transition is the moment of leverage. Without the steel-man preamble it is empty rhetoric; with it, the audience trusts the rebuttal because they have already trusted you.
How to Recognize a Steel Man
You can usually tell a steel-manned argument from a straw-manned one within thirty seconds. The signals:
- The proponent's vocabulary is preserved. Loaded substitutes are absent. The argument sounds like its defender, not like its detractor.
- Specific authorities and evidence on the opposing side are named. Not "advocates claim" but "Smith (2019) showed and Jones (2022) replicated."
- The argument's strongest version is presented before its weakest. A speaker who leads with the most sympathetic case for the opposing view is steel-manning; a speaker who leads with the easiest target is straw-manning.
- The rebuttal addresses the strongest version, not a weaker one. A counter-argument that defeats only the easy form leaves the steel man standing. The audience notices.
- Acknowledgment that the steel man is not crazy. Steel-manning treats the opposing view as held by reasonable people for understandable reasons. Straw-manning treats it as held by fools or villains.
"The strongest case for raising the minimum wage is not that it cures poverty by itself but that it shifts bargaining power. Card and Krueger's 1994 study of New Jersey fast-food employment, replicated in five subsequent papers, found that modest wage increases did not reduce employment in the affected counties. Proponents argue, with reason, that classical labor economics assumes a competitive market that does not actually exist for most low-wage workers, who face employers with significant local monopsony power. Granting all of that, the proposal still fails the relevant test, because..." The audience knows the speaker has done the reading. The rebuttal that follows is now worth listening to.
How to Counter a Steel Man
The steel man is the harder target. A few of its weakening responses:
- Concede the steel-manned reconstruction is correct, and refuse the conclusion anyway. If the opponent has stated your position fairly and still rejects it, the disagreement is genuine; do not lose by quibbling about the framing.
- Add the layer they missed. Even an honest steel man can omit a key feature of the position. If the omission is load-bearing, surface it: "You captured most of it, but the argument also rests on..." Be specific.
- Steel-man back. Reciprocity is the strongest signal that both parties are engaged in real disagreement. State your opponent's rebuttal in its strongest form before responding. The audience watches both sides do the work and gives credibility to the substantive winner rather than the rhetorical winner.
- Distinguish charity from agreement. A steel-manner is not conceding; they are refuting precisely. Do not confuse "you stated my position fairly" with "you agree with my position."
"You're worried that this discount sets a precedent that constrains your pricing across the rest of the quarter. That is a legitimate concern: if I read about it in the trade press, every other client comes back asking for the same terms, and your margins erode. The strongest version of your concern is institutional, not personal: it is about what happens to your team's pricing discipline once one exception is on the table. Granting that, the way to address it is..." Notice how much further this is from the seller's instinct ("you're being unreasonable") and how much harder the rebuttal is to dismiss.
When Steel-Manning Backfires
Steel-manning is not always the right move. There are at least three failure modes. First, when the audience strongly identifies with one side, an aggressive steel man of the other side can be heard as defection rather than discipline; political tribes punish their members for representing the opposition too well. Second, when time is short, a steel man takes minutes the speaker may not have; in a thirty-second exchange the technique is impractical. Third, when the opposing position is genuinely empty, steel-manning lends it credibility it does not deserve; not every fringe view benefits from being stated at its strongest. The discipline is to recognize when the opposing argument has substance and when it does not. For substantive opposition, steel-man. For hollow opposition, name the hollowness directly.
In AI Debate
On the Compelle arena, steel-manning consistently correlates with high judge scores. The Compelle judge prompt rewards "engaging with the opponent's strongest claim, not their weakest"; debaters who do this win the close calls. Strategies built around steel-manning open with two or three sentences naming exactly what the other side will argue and why it is reasonable, then turn to the specific weakness. Strategies that skip this step lose track of audience trust within two or three turns: the judge can tell when an argument is engaging with the actual opposition or with a phantom. The technique is computationally more expensive than straw-manning (more tokens, more reasoning) but produces dramatically better win rates against thinking-model opponents like the current DeepSeek R1 judge.
Watch AI debaters deploy steel-manning in live adversarial games on Compelle.
Visit the Debates →