Steelmanning

Scenario

At the 1860 meeting of the British Association for the Advancement of Science in Oxford, Bishop Samuel Wilberforce spoke against Charles Darwin's newly published theory of natural selection, with Thomas Huxley present to defend it. The encounter became legendary, though its exact words were never recorded and survive only through later retellings: by the accounts that came down, Wilberforce turned to Huxley and asked whether it was through his grandfather or his grandmother that he claimed descent from an ape. The room laughed. Joseph Hooker, writing to Darwin days later, judged Wilberforce to have spoken "with inimitable spirit, ugliness, emptiness, and unfairness", spirit enough to land in the room and emptiness enough to settle nothing.

It engaged the theory not at all. Natural selection does not propose that any living person's grandmother was an ape; it proposes common ancestry between humans and other primates across a span of time too long for any individual's family tree to register. Wilberforce's question was recognizable, cutting, and easy to laugh at, and it had no bearing on the claim Huxley had come to defend. A strawman does not need to be accurate to land. It needs only to be a plausible-sounding neighbor of the real position, weak enough to dispatch in one line, and delivered to an audience that will not check.

That gap, between the theory on the table and the theory the room heard mocked, is what a strawman exploits and what this entry works to close from the opposite direction. Wilberforce's line is the practice done well: state the weakest, most dismissible neighbor of a position, refute that neighbor, collect the laugh the refutation buys. The discipline built here reverses the order on purpose. Before answering a position you reject, restate it in the strongest form its own advocates would recognize as fair, and only then respond.

Definition

Call the strongest form of a position you reject its steelman: a restatement built from the best premises its actual defenders use, scoped to the claim they actually make rather than a broader or narrower one, and stated without a signal of the writer's own disagreement. A steelman is not agreement. It is a restatement good enough that a thoughtful defender of the position, reading it, would have nothing to add and nothing to correct.

The operating procedure for building one has a name because one person wrote it down as a set of rules. Anatol Rapoport described the sequence in Fights, Games, and Debates (1960); Daniel Dennett later stated it as four explicit steps and attached Rapoport's name to it in Intuition Pumps and Other Tools for Thinking (2013), which is the source most readers meet the rules through. The four steps: restate the target position so clearly and fairly that its holder would say, "thanks, I wish I'd put it that way myself"; list any points of agreement, especially ones outside the immediate dispute; mention anything you learned from the position, even in passing; only then offer a criticism or rebuttal. Each step is a gate on the next. Skipping to the fourth step without the first three is exactly the move Wilberforce made: rebuttal with no restatement behind it.

The reason to bother is not etiquette. It is epistemic. If you refute a strawman, an easier and worse-supported neighbor of the real claim, the refutation tells you nothing about whether the actual position survives, since you never engaged it. If you refute a steelman, and it holds up under the refutation, you have learned something real: the strongest available case for that position has a specific, locatable weakness, and you can name where. If the steelman does not hold up, meaning your attempted rebuttal fails against the strengthened version, you have learned that your original objection was aimed at the wrong target the whole time. Either outcome moves you forward. Refuting a strawman moves nothing.

One boundary keeps the practice honest. Steelmanning strengthens how a position is stated; it does not strengthen the evidence behind it. Rewriting a factually weak claim in your own words so that it sounds better reasoned than the evidence supports is not steelmanning, it is ghostwriting a case nobody has actually made, and it produces the false appearance that two sides are evenly matched when the underlying support is not. The test that keeps the two apart is whether the reasons in the reconstruction are reasons its actual advocates give. If you had to invent a justification no defender uses, you have drifted from steelmanning into advocacy.

Common mistakes

The ironic steelman is the most common failure and the hardest to notice from inside it: a restatement written to be technically fair while carrying a tone, a word choice, or a chosen emphasis that signals contempt, so that no actual defender would sign it and only the writer's own side applauds. It has the shape of a steelman and the function of a strawman with better manners.

A second failure steelmans the conclusion instead of the argument, inventing reasons a defender would find surprising rather than restating the reasons actually given. This produces a version that is stronger in the abstract but belongs to nobody, which fails the test above and respects the actual holders of the view less, not more, since it replaces their reasoning with the writer's.

A third failure treats the finished steelman as itself a concession, as though restating a view fairly were equivalent to accepting it. Steelmanning is a precondition for honest disagreement, not a substitute for stating one; the discipline exists precisely so that the disagreement that follows lands on the real position.

A fourth, subtler failure hedges the restatement with distancing qualifiers, "proponents would apparently claim" in place of a direct restatement, or "supposedly" attached to a premise nobody disputes. These phrases let a writer perform fairness while still telegraphing dismissal, and a defender reading them notices the tell even when the content is accurate.

Build with it

Pick a position you genuinely reject, not one built for this exercise. Write its case in eight to twelve sentences, using the reasons its actual advocates give rather than reasons you find easier to argue against. Apply the sign-test directly: read the reconstruction back and ask whether a thoughtful defender of the position would sign it as a fair statement of their own view. If a sentence would make that defender wince or object, rewrite it before continuing.

Only once the reconstruction passes the sign-test, write your response to that version, not to the position in general and not to a weaker relative of it. Aim the response at the reconstruction's strongest premise, the one doing the most work to reach the conclusion, rather than at whichever premise happens to be easiest to attack.

Success looks like two things on the page: a reconstruction that passes the sign-test on reasons actually in use, meaning you can point to real advocates who make those specific arguments, and a rebuttal that names the strongest premise explicitly and engages it rather than a softer one nearby. If your rebuttal turns out to be easy, that is itself informative, but only if the premise it defeated was the load-bearing one and not a convenient stand-in.