If the straw man is the easiest way to win an argument, the steelman is the most honest way to have one: restating the other side's view in its strongest form before responding. Far fewer can do it on demand, especially about an issue they care about. This level treats it as a skill with steps that can be practiced, tested and done badly.

Rapoport's rules

The philosopher Daniel Dennett, in a 2013 book on thinking tools, set out a procedure he credited to the game theorist Anatol Rapoport. In paraphrase, it has four steps. First, restate your opponent's position so clearly and fairly that they wish they had put it that way. Second, list any points on which you agree, especially ones that are not widely shared. Third, mention anything you have learned from them. Only then, fourth, are you permitted to say a word of rebuttal or criticism.

The order is the point. Criticism comes last, after the other person has heard you state their view well, seen what you share and heard what you learned. Dennett observed that opponents treated this way take criticism far better. The first step is also a test of yourself. If you cannot state the view well enough to satisfy someone who holds it, you are not yet ready to say it is wrong.

The ideological Turing test

In 2011 the economist Bryan Caplan proposed a sharper test. Alan Turing had suggested in 1950 that a machine could be called intelligent if a judge could not tell its written answers from a human's. Caplan adapted the idea: can you explain a rival view so well that its own supporters cannot tell you from one of them? He called this the ideological Turing test. Caplan was responding to a claim by the columnist Paul Krugman that one political side understood the other better than the reverse. Rather than argue about who understood whom, Caplan suggested measuring it.

The test is hard in a useful way. Most people can produce a version of the other side that their own friends would find convincing as a portrait. Producing one that the other side would sign is different, because it requires knowing what they actually find persuasive, not what you imagine they think.

How to build a steelman

Four habits help. First, find the view's best defenders, not its loudest. Every position has thoughtful people who have spent years on it; read or listen to them. Second, look for the value underneath. Most positions protect something their holders care about, such as safety, fairness, liberty or loyalty, and the steelman should name it. Third, ask what evidence the other side would cite and take it seriously. Fourth, and most important, check your version with someone who holds the view. A steelman that nobody on the other side recognizes is not a steelman. It is your own invention wearing their name.

Ways it goes wrong

Steelmanning has failure modes of its own. One is improving the view past what anyone holds, so that you end up arguing with a sophisticated version that no real person believes while ignoring the reasons people actually give. This can be its own kind of disrespect, since it implies their real reasons are not worth engaging. A second failure is using a steelman as a performance, a long display of fairness followed by a crushing rebuttal, which treats charity as a debating trick.

A third worry is that steelmanning gives bad ideas more credit than they deserve. Critics point out that some positions do not have a strong version, and that inventing one gives them false respectability. The reply is that steelmanning is not agreement. Stating a view at its strongest does not mean endorsing it; it means making sure that when you reject it, you are rejecting the real thing. Where a view really has no strong version, honest steelmanning will reveal that too.

Worked example: rent control

Jamal thinks rent control is a bad idea. He has read economists who argue that capping rents reduces the supply of housing and lets buildings decay. His first attempt to state the other side is: "People want rent control because they think landlords are evil." That is a straw man.

He tries again, following Rapoport. A stronger version reads: "Tenants who have lived in a neighborhood for decades face sudden rent increases that force them out of their schools, jobs and communities. Rent stabilization protects that stability, which markets do not price, and the harm of displacement falls on people with the least power to absorb it." He shows this to his neighbor Rosa, who supports rent control. She says it is close, but adds that for her it is also about older tenants who cannot move at all. He adds her point.

Jamal then lists what they agree on: that housing is too expensive and that displacement is a real harm. He notes that Rosa taught him something he had not considered about elderly tenants. Only then does he lay out his concern about housing supply. Rosa still disagrees, but she listens to his argument in full, because he has shown that he understands hers.