Steelmanning is the practice of engaging with the strongest possible version of an opposing argument — the opposite of strawmanning. Instead of attacking the weakest formulation of a position (which proves nothing), you reconstruct the argument in its most compelling form and then evaluate that. Steelmanning serves both epistemics and discourse: epistemically, it protects against confirmation bias by forcing genuine engagement with counterevidence. Discursively, it builds trust and surfaces real disagreements rather than misunderstandings. The practice requires understanding the opposing position well enough to argue for it convincingly — which often reveals considerations you had overlooked.
Choose a position you disagree with and write the best argument FOR it that you can. Show it to someone who holds that position and ask if you represented them fairly. If they say no, revise until they agree. Notice what you learn in the process — steelmanning almost always reveals something you had not considered.
From considering the opposite, you know that deliberately generating reasons for the opposing conclusion is one of the most effective debiasing techniques. Steelmanning takes this a step further: instead of just listing counterarguments, you reconstruct the opposing position in its strongest possible form -- the version that a smart, well-informed advocate would actually endorse -- and then engage with that version rather than a weakened caricature.
The term is the opposite of strawmanning, in which you attack the weakest or most distorted version of an opposing argument. Strawmanning is rhetorically effective (it is easy to demolish a bad version of a position) but epistemically worthless (demolishing a position no one holds proves nothing about the real position). Steelmanning reverses this: you construct the version of the argument that is hardest to refute. The test of whether you have done it correctly is whether someone who actually holds the position would read your reconstruction and say "yes, that is what I believe -- you have represented me fairly." If they say "no, you are missing the point," you have not steelmanned; you have just constructed a different kind of distortion.
The epistemic payoff of steelmanning is that it almost always reveals considerations you had overlooked. If you disagree with a position and you can already see all the reasons someone might hold it, the exercise is trivial -- which means it almost never is. The process of genuinely understanding why a thoughtful person holds a position you reject forces you to engage with evidence, arguments, and values that your confirmation bias had filtered out. Sometimes you discover that the opposing position is weaker than you expected, and your original view is confirmed through genuine scrutiny rather than lazy dismissal. Sometimes you discover that the opposing position has a stronger foundation than you realized, and you need to adjust your view. Either outcome is epistemically valuable; what is not valuable is maintaining a disagreement based on a misunderstanding of what the other side actually claims.
Steelmanning also has a discursive benefit that goes beyond personal epistemics. In any conversation or debate, people can tell whether you have genuinely understood their position or are attacking a caricature. When you demonstrate accurate understanding -- when you can articulate their view better than most of their allies can -- you earn trust and create the conditions for productive disagreement. Real disagreements, where both parties understand each other's actual claims and still differ, are rare and valuable. Manufactured disagreements, where both parties are attacking distorted versions of each other's positions, are common and worthless. Steelmanning is the practice that separates the two, and it is why the Rationalist tradition treats it as a foundational epistemic skill rather than just a debating courtesy.
Topics in reflective domains aren't scored by quiz answers. Read, reflect, and mark when you've thought it through.