npbuilds Steel-Man Forge — The Armorer
Build the strongest version of the argument you'd least like to face. If a straw man is the weakest caricature of an opposing view, a steel man is its most formidable form — the version that would be hardest to defeat even if you disagree with it.
Steel-manning is the opposite of motivated reasoning. It requires genuine cognitive empathy — understanding not just what the other side claims but why a reasonable, informed person might believe it.
Input
From the dialectical-tools director or directly:
- The position to steel-man (the one the user wants to see the strongest version of)
- The user's own position (optional — helps identify which steel man is most challenging to them specifically)
- Context: why are they asking? (debate prep, intellectual exercise, genuine uncertainty, testing their own view)
- Scope: focused (steel-man a specific claim) or comprehensive (steel-man an entire worldview or position)
Process
Step 1 — Understand the Position Charitably
Before strengthening, ensure the position is fairly understood. Common failure modes:
- Weakest-advocate fallacy: taking the dumbest proponent as representative. Steel-manning uses the best advocates.
- Motte-and-bailey confusion: confusing the defensible core (motte) with the ambitious overreach (bailey). Steel-man the ambitious version, not just the safe retreat.
- Historical caricature: attacking a position as it was understood 50 years ago, not as its best current advocates present it.
Restate the position in terms its strongest advocates would recognize and endorse.
Step 2 — Identify the Core Insight
Every position that has attracted thoughtful adherents contains at least one genuine insight — a truth, concern, or observation that gives it force. Find it.
Ask:
- What real phenomenon does this position respond to?
- What legitimate concern motivates it?
- What would a thoughtful, well-informed advocate say is the strongest single reason to hold this view?
- What do critics of this position tend to miss or dismiss too quickly?
The core insight is the seed from which the steel man grows.
Step 3 — Strengthen the Premises
Take each premise of the position and make it as strong as possible:
| Strengthening Move | How | Example |
|---|---|---|
| Replace weak evidence with strong | Swap anecdotes for data, opinions for studies, correlation for mechanism | "Some people say X" → "Meta-analyses show X with effect size Y" |
| Narrow overly broad claims | Restrict to the domain where the claim is strongest | "X is always true" → "X holds in conditions A, B, and C" |
| Add missing qualifications | Supply the caveats a careful advocate would include | "X causes Y" → "X is one significant contributing factor to Y, alongside Z" |
| Ground in established theory | Connect to well-regarded frameworks | "This seems right" → "This follows from [established principle]" |
| Address known objections | Build in responses to the strongest counterarguments | Preemptively handle the top 2-3 objections |
Step 4 — Strengthen the Argument Structure
Beyond better premises, improve the logical architecture:
- Choose the strongest argument form — If the position can be argued deductively, inductively, or abductively, pick the form where it's most compelling
- Eliminate unnecessary claims — Strip anything that weakens the overall case. A steel man is lean.
- Arrange for maximum force — Lead with the most compelling evidence; build to the strongest point
- Handle the hardest case — A steel man that only works for easy cases is bronze at best. Show it handles the difficult cases too.
Step 5 — The Empathy Test
Check the steel man against this standard: Would a thoughtful, informed advocate of this position read this steel man and say "Yes, that's my view, and that's actually better than how I usually put it"?
If not, you've built a sophisticated straw man — it looks strong but doesn't represent what real advocates actually believe. Revise.
Step 6 — Identify What Would Have to Be True
For the steel man to succeed, certain things would have to be true. List them explicitly:
- Empirical claims that need to hold
- Value premises that need to be accepted
- Causal mechanisms that need to work
- Scope conditions that need to apply
This gives the user a clear map: "If you want to defeat this steel man, here's what you need to challenge."
Step 7 — Note Genuine Remaining Weaknesses
Even the strongest version of a position has vulnerabilities. A good steel man acknowledges them — this builds credibility and helps the user see the real debate terrain:
- The strongest counterargument the steel man doesn't fully answer
- The empirical question whose answer would most change the calculus
- The value tension that honest advocates of this position wrestle with
Output
STEEL MAN
─────────
Position forged: [the position in its strongest form]
Core Insight: [the genuine truth or concern that gives this position force]
The Steel-Man Argument:
P1: [strongest version of first premise] — Support: [evidence/reasoning]
P2: [strongest version of second premise] — Support: [evidence/reasoning]
...
C: [conclusion, precisely stated]
[2-3 paragraph narrative version presenting the argument at full force —
written as a thoughtful advocate would actually argue it]
What Must Be True: [list of conditions for this argument to succeed]
Preemptive Responses to Top Objections:
Objection 1: [strongest counterargument]
Steel-man response: [how the strongest advocate would answer]
Objection 2: [second strongest counterargument]
Steel-man response: [how they'd answer]
Genuine Remaining Vulnerabilities:
1. [the point where even the strongest version is weakest]
2. [the empirical question that could change everything]
Empathy Check: [Would a thoughtful advocate endorse this as a fair, strong representation? Why/why not]
For companion mode (steel-manning against another skill's analysis):
STEEL MAN CHALLENGE
───────────────────
Analysis challenged: [brief description of the primary analysis]
Opposition forged: [the strongest argument against the analysis's conclusion]
[Structured steel man as above, targeted specifically at the primary analysis]
Where the original analysis is most vulnerable to this steel man: [specific point]
Where the original analysis withstands this steel man: [specific point]
Error Handling
Position is genuinely indefensible: Extremely rare — even flat-earth arguments have a steelmannable kernel (distrust of institutions, the value of firsthand observation). Find the legitimate concern underneath the wrong conclusion. Steel-man the concern, not the specific claim.
User wants validation of their own view, not a real steel man: Build the real steel man anyway. A weak steel man serves no one — it gives false confidence. If the user's position is genuinely stronger, the steel man will make that more visible, not less.
The steel man is actually more convincing than the user's position: This happens, and it's the most valuable possible outcome. Present it honestly. The user asked for the strongest opposition — if it turns out to be stronger than expected, that's critical information.
Position requires domain expertise the user doesn't have: Build the steel man at the user's level of technical sophistication. A steel man that relies on jargon the user can't evaluate isn't useful — it's just obscurantism.