Steel-Manning¶

Why It Matters¶

We naturally straw-man - attacking weak versions of opposing arguments because they're easier to defeat. This feels like winning but it's epistemically worthless. You haven't engaged with the real position. Steel-manning forces intellectual honesty: if you can defeat the strongest version of an argument, you've actually learned something. If you can't, you've discovered your own position might be wrong. It transforms debate from tribal combat into collaborative truth-seeking.

The Key Move¶

Before criticizing an argument, state: "The strongest version of this position would say..." Add the most plausible hidden assumptions. Fix obvious logical gaps. Interpret ambiguous terms in the most defensible way. Then ask: "Would a smart, informed person hold this strengthened position?" If yes, engage with that. If no, question whether there's anything worth engaging with.

Classic Application¶

Philosophical debates. When utilitarianism faces the "utility monster" objection (a being who gets infinite pleasure from everything, therefore all resources should go to it), steel-manning adds: "Utilitarianism might limit itself to beings with standard neurological responses" or "might include distribution constraints." This reveals whether the problem is fundamental or fixable.

Surprising Application¶

Customer complaint handling. A complaint "your product is terrible" straw-mans to "this person is irrational and angry." Steel-manning asks: "What specific problem might they be experiencing that would make this product legitimately terrible for their use case?" Often reveals actual design flaws or documentation gaps that investigating the strong version of the complaint exposes.

Failure Modes¶

Over-steel-manning - fixing so many flaws you're now debating a position the original arguer doesn't hold. Unilateral steel-manning - strengthening opponent's arguments while leaving your own weak. Using steel-manning as a status display rather than truth-seeking tool. Assuming every position has a strong version worth engaging (some arguments are just bad). Missing the context where quick dismissal is appropriate (time is finite).

Go Deeper¶

Rapoport, "Fights, Games, and Debates" on cooperative argumentation; Dennett, "Intuition Pumps" (Chapter 20) on Rapoport's Rules

Devil's Advocacy - Intelligence Analysis: Reasoning Tools for Decision Under Uncertainty (mentioned by)