The Steelman Method

How we evaluate our own claims

What Steelmanning Is (And Isn't)

Steelmanning is the opposite of strawmanning. Instead of weakening an argument to knock it down, you strengthen it to its most defensible form before evaluating.

When applied to your own work, this means:

  • Finding the strongest version of each claim
  • Identifying exactly where each claim's support comes from
  • Making explicit what would falsify it
  • Distinguishing what's locked vs. what's optional

It's not about defending claims. It's about understanding them clearly enough to know which ones deserve defense.

The Five-Step Process

Step 1: Isolate the Claim

State the claim precisely. Not "RSM is interesting" but "RSM derives that O1 is unoccupiable from Postulate 1 and Theorem 0.5."

Ask: What exactly is being claimed? Can you state it in one sentence without hedging?

Step 2: Trace the Dependencies

Follow the claim back to its foundations. What axioms does it require? What postulates? What previous theorems?

Ask: What would have to be false for this to be false? Draw the dependency chain.

Step 3: Strengthen to Defensible Form

Find the version of the claim that actually follows from the framework. Often this means narrowing scope or adding conditions.

Ask: Is there a more precise way to state this that loses less to objections?

Step 4: Identify the Gap

Find what's missing. Every claim has limits. What doesn't follow? What's assumed? What's the weakest link?

Ask: Where is this claim most vulnerable? What would a careful critic attack?

Step 5: Classify Epistemic Status

Place the claim in its proper tier:

  • Tier 1 (Locked): Follows necessarily from framework
  • Tier 2 (Postulate-Dependent): Follows given optional assumptions
  • Tier 3 (Empirical): Testable prediction
  • Tier 4 (Analogical): Pattern observation, not derivation
  • Tier 5 (Outside Scope): Not addressed by framework

Worked Example: "RSM explains consciousness"

Step 1 - Isolate:

"RSM's framework applies to conscious experience, explaining why there's 'something it's like' to be aware."

Step 2 - Dependencies:

This would require: (a) consciousness to be a contrast-based representational system, (b) the framework to address experiential qualities, not just structural patterns.

Step 3 - Strengthen:

Best defensible version: "Conscious experience exhibits structural features (irreducible contrast, dynamic oscillation, recursive self-reference) that parallel RSM's three requirements."

Step 4 - Gap:

The gap: Structural parallel doesn't explain why there's experience at all. RSM describes patterns; the hard problem is about existence of qualia. Pattern recognition doesn't bridge this gap.

Step 5 - Classify:

Tier 5 (Outside Scope) - The consciousness claim isn't derivable, testable, or even analogical in a strong sense. RSM deliberately excludes experiential claims from its scope.

Result: The steelman version of "RSM explains consciousness" is "RSM notices structural parallels in consciousness." That's a much weaker but honest claim. The original overreaches.

Why This Matters

Frameworks like RSM are easy to over-inflate. Pattern recognition is seductive. Finding the same shape in multiple places can feel like discovering a cosmic truth.

The steelman method is a reality check:

  • For the creators: Keeps us honest about what we've actually shown vs. what we'd like to claim
  • For the readers: Provides explicit criteria for evaluating claims
  • For the framework: Strengthens what survives, removes what doesn't

A framework that can't survive its own steelman analysis wasn't worth defending.

Try It Yourself

Pick any RSM claim that interests or bothers you:

  1. State it precisely
  2. Trace what it depends on
  3. Find its strongest form
  4. Identify where it's weakest
  5. Classify its epistemic status

If you find a claim that doesn't survive this process, tell us. That's how frameworks improve.

The Underlying Commitment

RSM is developed with this principle:

"We would rather discover our claims are wrong than continue believing they're right. The framework improves through honest critique, not defensive entrenchment."

If you can break something in RSM, you've done the project a service. Claims that can't be broken honestly weren't strong to begin with.