Theme
βWhat is a steelman?β
Every argument on this site starts with its steelman: the case as its best advocates would make it. This page explains what that means, where the idea comes from, how the site tests its own steelmen, and what a steelman is not.
The picture
The opposite of a strawman.
A strawman is a weakened copy of someone's view, built to be knocked down. A steelman is the reverse: the view restated in its strongest form, with its best evidence, in words its holders would recognise as their own. On this site that is the section 'Why people find it convincing' at the top of each argument. It comes before anything is said against the argument, and it is written in good faith: the facts that make the argument feel right are stated as facts when they are true.
An old idea with a new name.
The word is recent internet slang, but the practice is old. John Stuart Mill put it in 1859: 'He who knows only his own side of the case, knows little of that.' Hearing the other side from your own teachers is not enough: '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.' Daniel Dennett set out 'Rapoport's rules' (after Anatol Rapoport) in Intuition Pumps (2013): 'You should attempt to re-express your target position so clearly, vividly, and fairly that your target says "Thanks, I wish I'd thought of putting it that way"'; then list points of agreement and 'anything you have learnt from your target'; 'Only then are you permitted to say so much as a word of rebuttal or criticism'. He admitted it is hard: 'Some targets, quite frankly, don't deserve such respectful attention, and β I admit β it can be sheer joy to skewer and roast them' (both as quoted by Cass Sunstein, 2025). The economist Bryan Caplan proposed an 'ideological Turing test' (2011): 'the ability to pass ideological Turing tests β to state opposing views as clearly and persuasively as their proponents β is a genuine symptom of objectivity and wisdom'; in his experiment, observers try to tell the real believers from the people imitating them.
primary documentJohn Stuart Mill, On Liberty (1859), chapter 2 (1859) β
Chapter II, 'Of the Liberty of Thought and Discussion'; the passage 'He who knows only his own side of the case, knows little of that β¦' read in the Project Gutenberg text. Re-read 2026-10-12 (Gutenberg text): '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'.
advocacy / viewDaniel C. Dennett, Intuition Pumps and Other Tools for Thinking (2013), 'Rapoport's rules', as quoted by Cass Sunstein, 'The Rapoport Rules' (Substack, 24 April 2025) (2013) β
Read 2026-10-01 (the book itself not read). Sunstein quotes the four rules and Dennett's admission ('Following Rapoport's Rules is always, for me at least, something of a struggle β¦'), credits Dennett's 2013 book via Riccardo Rebonato, and calls Rapoport 'a mathematical psychologist'. Re-read 2026-10-12 (Sunstein's post): the rules include listing 'anything you have learnt from your target' and 'Only then are you permitted to say so much as a word of rebuttal or criticism'. Re-read 2026-10-13 on Sunstein's post: rule 1, 'You should attempt to re-express your target position so clearly, vividly, and fairly that your target says βThanks, I wish I'd thought of putting it that way.β'; and Dennett: 'Some targets, quite frankly, don't deserve such respectful attention, and β I admit β it can be sheer joy to skewer and roast them.'
advocacy / viewBryan Caplan, 'The Ideological Turing Test' (EconLog, 2011) (2011) β
EconLog, 20 June 2011. Re-read 2026-10-12: 'the ability to pass ideological Turing tests β to state opposing views as clearly and persuasively as their proponents β is a genuine symptom of objectivity and wisdom'.
How this site tests its own steelmen.
Three checks. First, the signature test: would a thoughtful advocate of this argument sign the steelman as a fair statement of their case? If not, it is rewritten before anything else. Second, a search for the current leading advocates and their most-cited texts, because the first version that comes to mind is usually the most-written-about one, not the strongest (see 'You need the argument to find the argument'). Third, anchoring: each argument lists where it has actually been made in public, by whom and in which words ('Where it's been said'), so the steelman stays tied to what people really argue.
What a steelman is not.
It is not agreement: stating a case at full strength is how you find out exactly where it holds and where it breaks. It is not the most extreme version: the strongest case is usually the most careful one, not the loudest. And it is not an improved argument that nobody makes: an argument is only on this site if people actually make it in public; a clever version invented here would be a strawman in reverse, and it would teach readers to answer a case that nobody holds.
Why it matters here.
In this conflict each side mostly meets the other's arguments in the form its own side has chosen to show β often the weakest or ugliest version, which makes the other side look stupid or evil. People who then meet the real argument for the first time are unprepared, and often swing all the way over. A steelman works the other way round: it shows the case at its best, then what is true in it, what it leaves out and where it leads. That is also why this site states propaganda arguments at full strength (see 'Honesty and inoculation').
What it looks like when it works: a documented case.
Megan Phelps-Roper grew up in the Westboro Baptist Church, known for its pickets against gay people and Jews, and left it in 2012. In a 2017 talk she described how: people she argued with on Twitter answered her with civil curiosity instead of contempt. 'A man named David' β a Jewish blogger in Jerusalem β 'after several months of heated but friendly arguments online β¦ came out to see me at a picket in New Orleans. He brought me a Middle Eastern dessert from Jerusalem', and 'We'd started to see each other as human beings, and it changed the way we spoke to one another.' She drew four steps from it: 'don't assume bad intent'; 'ask questions'; 'stay calm'; and 'make the argument' β because 'we sometimes assume that the value of our position is or should be obvious and self-evident', and 'if they hadn't actually made their arguments, it would've been so much harder for me to see the world in a different way'. None of them is agreement; all of them keep the conversation open long enough for reality to do its work. That is the reader this site is written for.
primary documentMegan Phelps-Roper, 'I grew up in the Westboro Baptist Church. Here's why I left' (TED, February 2017), transcript (2017) incident β
Transcript read: the David episode and the four steps ('don't assume bad intent', 'ask questions', 'stay calm', 'make the argument'). The transcript page now renders by script (13 Oct 2026), so the quoted lines were not re-read.
Where it leads if pressed
Working it out: does this site pass its own test?
The question: this page sets three checks for the site's steelmen. How many of the 59 argument pages pass each, on the site's own data? (1) Anchoring, the one check that can be counted: 21 argument pages show the argument being made in public by two or more different speakers, and 36 by one (the site used to label both 'Common argument', the second with 'e.g.'; since 12 October 2026 only the 5 pages with three or more different speakers are labelled 'Common argument', and the rest show 'Made publicly by' with the names), one is anchored only by someone else's report of it, and one not at all. By the side making the argument, 9 pro-Israeli and 8 pro-Palestinian arguments have two or more instances, and 13 and 14 have one; the rest have no side recorded (counts from the pages' 'Where it's been said' entries, October 2026). So most arguments here rest on one public instance: enough to show that someone makes the argument, not that it is common. (2) The signature test: it has been applied by the AI model that drafts the pages and by the editor, not by advocates of the arguments; no page records an advocate reading its steelman. As run, it is a self-check, the kind 'Limits' below says a writer can bend. (3) The search for leading advocates: a working rule since September 2026, but its results are not shown on the pages, so a reader cannot see which steelmen went through it. What's at stake: for readers on either side, a steelman checked only by its writer can still be a quiet strawman. The anchoring counts are close to even by side, which is evidence against a tilt in which arguments get anchored, but says nothing about how each is stated. Holds on the site's own data: all but two arguments are tied to at least one named public instance, and anchoring is about as thorough for each side's arguments. Open, needs a test by advocates: that the people who make these arguments would sign the steelmen. Not supported: that the signature test has been passed in the sense this page describes, by advocates rather than by the writer.
Limits.
Whoever writes a steelman also decides what 'strongest' means, and can pick a version that is easier to answer β the check against that is the advocates' own words, quoted and linked. And for some readers, stating a hateful argument at full strength will feel like giving it a platform; the answer this site gives is that the argument is already out there, and it works best when nobody has looked at it closely.