| name | blind-steelman |
| description | Use when disagreeing with a position, dismissing an alternative approach, writing a comparison or decision record, or when an opposing view seems obviously wrong. Applies steelmanning and Bryan Caplan's Ideological Turing Test — you understand a position only when its holders would accept your statement of it. |
Steelman / Ideological Turing Test
Overview
The inverse of strawmanning: before rejecting a position, construct its strongest version — stronger than its actual advocates managed, if possible. The test of success is Bryan Caplan's Ideological Turing Test (2011): state the opposing position such that its genuine holders couldn't distinguish you from one of their own. If you can't pass, you don't yet understand what you're rejecting — you're rejecting your own compressed sketch of it, and whatever the sketch dropped is your blind spot.
"Obviously wrong" usually means "compressed until wrong." The lossy compression is yours.
When to Use
- Dismissing an alternative: library, architecture, strategy, interpretation ("Redux is overkill", "they just don't get it")
- Writing decision records / trade-off comparisons (the rejected option deserves its best case on record)
- Any disagreement where the other side contains smart people — the disagreement itself is evidence you're missing their inputs
- Reviewing your own past decision that now looks obviously right
The Procedure
- Write the position as its smartest advocate would — their strongest arguments, their evidence, the values and constraints that make it rational. No "some people think…"; write it in first person, from inside.
- Run the Turing test: would an actual proponent sign off on your rendering? If reachable, literally ask. If not, compare against the best published advocacy — did you include their best points or only the quotable weak ones?
- Locate the crux: with the steelman standing, find the specific premise where you actually diverge (a fact? a value weight? a predicted outcome?). Disagreements shrink to cruxes; blind spots live in the difference between the crux and what you originally attacked.
- Then judge. Rejecting a steelman you built is informative. Rejecting a strawman is noise.
Quick Reference
| Tell | Fix |
|---|
| "They just want / they just don't understand…" | No "just" — write their full model |
| Can state your side's evidence but not theirs | You have a conclusion, not an understanding |
| Rejected option gets 1 line in the ADR | Give it its best paragraph before rejecting |
| Their position seems to have no upside | You've compressed it to caricature — decompress |
Common Mistakes
- Steelmanning into a position they don't hold ("what you SHOULD be arguing is…") — that fails the Turing test too; fidelity first, charity second.
- Performing the steelman then judging on the original strawman anyway. The verdict must cite the steelman's crux.
- Reserving steelmanning for opponents' views only. Steelman the option YOU rejected last month before building on that rejection.