Steel-man

What it is

Imagine the person who most wants the option you are about to turn down. They read your description of it. Would they say "yes, that is my argument"? If so, you have built a steel man: the strongest honest version of an option you plan to reject.

You build it before you reject anything. It has three parts: the goal that option really serves, its best evidence and what would have to be true for it to win. This is not just about fairness. Your choice has to beat the best rival, not a weak description of it.

Where it comes from

The word has no single author. It was made as the opposite of the much older "straw man": a weak version of someone's argument, easy to knock down. The earliest example in the Word Spy dictionary is a 2003 forum post. Luke Muehlhauser defined it for debate and made it popular. On 24 October 2011, writing as lukeprog, he published "Better Disagreement" on the website LessWrong. In a footnote he defined a steel man as "an improvement of someone's position or argument that is harder to defeat than their originally stated position or argument." In December 2012 Chana Messinger took the idea further in "Knocking Down a Steel Man", on her blog The Merely Real.

The habit is older than the word. In On Liberty (1859) the philosopher John Stuart Mill wrote that someone who knows only his own side of an argument knows little of it. Of the other side's arguments he wrote: "He must be able to hear them from persons who actually believe them; who defend them in earnest, and do their very utmost for them. He must know them in their most plausible and persuasive form." The game theorist Anatol Rapoport wrote rules for criticism. The first is to restate the other view "so clearly, vividly, and fairly that your target says, 'Thanks, I wish I'd thought of putting it that way'". The philosopher Daniel Dennett reprints the rules in Intuition Pumps and Other Tools for Thinking (2013). Philosophers have a related idea, the principle of charity: read an argument in its most sensible form. Neil L. Wilson named it in 1958–59. Willard Quine and Donald Davidson developed it.

What it corrects

In a recommendation you describe the options you turn down in your own words. You judge them by your goal. But each was built for a different goal, so it loses before anyone looks closely. Even honest people make this mistake.

Nothing here feels like cheating, so careful people miss it too. You check the sums, but the comparison was unfair from the start. And the mistake helps you: the weaker the rival looks, the surer your conclusion sounds.

How it works

1. Name the option you plan to reject. If there are several, pick the strongest, not the easiest to beat.
2. Write its goal in its supporters' words. What do they want to protect?
3. Give it its best two or three facts, from the sources you used for yours.
4. Write what would have to be true for it to win.
5. Check that against what you know, and say how you know it.
6. Argue your option against this version. Write down what changed.

Worked example

Charles Darwin argued against himself in print. Chapter VI of On the Origin of Species (1859), "Difficulties on Theory", opens: "Long before having arrived at this part of my work, a crowd of difficulties will have occurred to the reader. Some of them are so grave that to this day I can never reflect on them without being staggered." He then sets out the objections at full strength, among them the missing in-between forms and the eye's origin.

Under attack, he kept the habit. In 1871 the biologist St George Jackson Mivart published On the Genesis of Species. Mivart argued that natural selection could not explain the early stages of a useful body part. Half an eye sees nothing, so why would it survive? Darwin did not restate the objection in an easier form. In the sixth edition, finished in January 1872, he added a new Chapter VII: "Miscellaneous Objections to the Theory of Natural Selection". Most of it is about those early stages. Because he kept the objection in Mivart's terms, he had to answer it precisely. His finished theory carries the marks of that answer.

In a Business Case Weekly case

In the Rolex case, a question under "Scarcity Meets Its Limits, 2015-2022" is set in 2021. On the resale market, Rolex watches sell for roughly three times the shop price. Rolex could make about 30 per cent more watches, raise shop prices by about 30 per cent or start a programme to certify second-hand watches. You are asked to argue for the option Rolex rejected most firmly.

Without a steel man, it is easy to call one option greedy and another timid. In fact each has a real goal, a sensible explanation of the price gap and a situation where it would be right. Build all three at full strength. Only then choose, and say what you give up.

In your answer

• "The strongest argument for the option I reject is that it protects …, and its best evidence is …"
• "Its supporters would say the goal is not … but …"
• "That argument wins if … is true. Here it is not, because the case reports …"
• "Building that steel man changed one thing in my answer: …"

Common misuse

The common mistake is a polite paragraph. You give a fair-sounding account of the rival, then write the conclusion you would have written anyway. It looks like a steel man but changes nothing. To check, ask what the steel man changed in your recommendation. A number? A condition? How sure you are? When you would revisit the decision? If nothing changed, and you cannot name evidence that would have changed it, you complimented the rival instead of comparing.

The rarer mistake is the reverse. You improve the rival into a version nobody in the room holds. You defeat it, and the real objection still stands. Stick to the version a real supporter would sign.

References

• John Stuart Mill, On Liberty, chapter 2, "Of the Liberty of Thought and Discussion" (1859). Why you should know the other side at its strongest. An evening's read.
• lukeprog (Luke Muehlhauser), "Better Disagreement", LessWrong, 24 October 2011. The post that made the term popular, with the definition in a footnote.
• Charles Darwin, On the Origin of Species, chapter VI, "Difficulties on Theory" (1859). The worked example, in Darwin's own words.
• Daniel C. Dennett, Intuition Pumps and Other Tools for Thinking, W. W. Norton, 2013, chapter 3, on Rapoport's rules for criticism. No link: no authorised full text is online.