Steelman and Charitable Precision

By the end of this lesson, you can reconstruct the strongest faithful version of a claim, ask its speaker to correct it, and then criticize a specific part without changing the position first.

The core idea: A steelman is not the strongest argument you can invent nearby. It is the strongest version the speaker can still recognize as theirs.

Core Insight

In the previous lesson, Mina and Daniel separated what Mina said from what Daniel first inferred. They are considering a proposed rule for a phone-free discussion group: try it for two visible meetings, then decide whether to keep it. Mina says: “A short visible trial can show whether participants accept a rule, but it may not show whether ordinary attention changes after the observation ends.”

Daniel answers quickly: “So you reject experiments because you demand perfect data.” His sentence is easy to attack. Mina did not reject experiments, and she did not demand perfection. She questioned whether one particular measurement could support one particular conclusion. Daniel may now win an argument against a convenient substitute while leaving Mina’s actual concern untouched.

Consider the more demanding move. Daniel pauses and offers this reconstruction: “You are not saying the trial is useless. You are saying we should not treat two observed meetings as evidence that attention will stay different once the trial is over, because observation can change behavior. Is that your concern?” Mina replies: “Mostly. I would also say the trial can tell us something about acceptance, even if not about durability.”

That exchange produces a usable object of disagreement. Daniel has not agreed with Mina. He may still think a low-cost trial is good enough to justify trying the rule. But he now knows what must be answered: whether evidence of acceptance is sufficient for a reversible policy, or whether the group needs evidence about lasting attention before treating the rule as successful.

The term steelman names this disciplined reconstruction. Informal-logic accounts of charity advise interpreters to choose the most credible reading compatible with the text and, when a missing premise is necessary, use the logically minimum addition rather than a grander replacement. Stanford Encyclopedia of Philosophy, “Informal Logic” A steelman makes that advice conversational: preserve the claim, reasons, limits, and stakes, then let the speaker say whether you got it right.

There is a real trade-off. This practice takes time and can make a reply harder to defeat. It can also go wrong if “charity” erases a harmful effect, removes the people affected, or quietly turns an ugly claim into a nicer one. Precision is not politeness theater. You do not have to agree, excuse harm, or manufacture symmetry between positions. You have to be clear about what you are criticizing.

The Faithful-Steelman Procedure

Treat your first paraphrase as a hypothesis, not a verdict. The procedure below is deliberately small enough to use in a meeting, a seminar, or a tense message thread.

1. Preserve the original before improving anything

Write down the claim in the speaker’s own words, or quote it closely. This is your control version. In Mina’s case:

“A short visible trial can show whether participants accept a rule, but it may not show whether ordinary attention changes after the observation ends.”

Without this step, it is easy to confuse your polished summary with what was actually said. If the original is vague, preserve the vagueness and ask a question; do not silently resolve it in your preferred direction.

2. Separate conclusion, reasons, and scope

Name three things in plain language:

• Conclusion: Do not treat two visible meetings as enough evidence of lasting attention change.
• Reasons offered: Observation can change behavior; the decision concerns attention after observation ends.
• Scope or qualifier: The trial may still teach the group about acceptance or preference; it is not useless.

This is not yet a formal argument map. You are simply making its load-bearing pieces visible. A conclusion without its qualifier becomes too broad; reasons without a conclusion become a list of worries.

3. Add only the minimum necessary link—and label it

Sometimes a speaker has supplied reasons but not the bridge to the conclusion. You may add the smallest bridge needed to make the reasoning intelligible, but mark it as an addition rather than smuggling it in as their view.

For Mina, the bridge might be: “Evidence gathered under observation may not generalize to ordinary conditions.” It is a plausible link, but Daniel should say, “I think the link you are relying on is…” and invite correction. He should not add a much stronger thesis such as “All experiments are artificial and therefore worthless.” That thesis changes both the scope and the evidence burden.

4. Return the reconstruction for a fidelity check

Use a question that gives the other person room to reject, narrow, or revise it:

“Would you accept this as your position: the trial can test acceptance, but it cannot by itself show durable attention change, so we should not claim more than it measures?”

Mina’s “Mostly” matters. Her correction adds a boundary: the issue is not whether to run a trial but what conclusion to draw from it. Record the correction. A steelman becomes faithful through this feedback loop, not through your confidence that you were being fair.

5. Critique a named target

Only now decide what you dispute. Daniel can say: “I accept that the trial does not establish durability. My objection is narrower: because the rule is reversible and acceptance is one decision-relevant outcome, evidence of acceptance may be enough to justify a next trial. We should just avoid calling it proof of lasting attention.”

This response has a target: the standard of evidence needed for a reversible next step. It does not pretend Mina claimed experiments are bad. Mina can answer that acceptance alone may be insufficient if the rule imposes costs, but now both people are working on the same decision.

Three Reconstructions, Only One Faithful

Compare Daniel’s possible replies. Each sounds more sophisticated than a direct insult, which is exactly why the distinction matters.

The unauthorized improvement is a common trap. It is kinder than the strawman, but kindness alone is not fidelity. Imagine Mina had merely said, “We should not overclaim.” Transforming that into a demand for a formal study may make her sound admirable and make the debate more interesting, but it gives Daniel a different opponent.

Check yourself. A student says, “I do not think the book club should choose every book by simple majority; a stable minority may never get a pick.” Which reply is a steelman?

1. “You hate democracy and want a committee to dictate everyone’s reading.”
2. “You claim voting is always oppressive, so the club must abandon it entirely.”
3. “You are concerned that a repeated majority rule can exclude recurring minority preferences; you may be asking for a rule that protects some rotation.”

Choose before reading on. The answer is 3, because it preserves the stated problem and marks the proposed remedy as provisional. It does not attribute hostility to democracy or an all-or-nothing abolition of voting.

Charity Has Boundaries

“Interpret the best version” does not mean “make every position reasonable.” It means distinguish accurately between these tasks:

• Reconstruction: What is the person actually asserting, and what reasons do they give?
• Evaluation: Are the reasons true, relevant, sufficient, or ethically acceptable?
• Decision: Given the effects and risks, what should we do?

Keeping them apart matters most when a claim concerns harm or unequal power. Suppose a manager says, “We should publish employees’ productivity rankings because competition motivates people.” A faithful reconstruction must retain that rankings expose employees to comparison and can change power dynamics; it must not sanitize the proposal into “The manager merely wants everyone to flourish.” You can then criticize the evidence, the distribution of costs, the consent problem, or the policy itself. Fidelity makes that critique sharper because it names the actual mechanism and affected people.

Likewise, a speaker may be wrong about facts. Steelmanning a false medical claim does not require presenting it as medically credible. You might say: “I hear that you are worried about side effects and distrust the source of the recommendation. Those are the concerns we should examine; the claim that the treatment contains ingredient X is still factually testable.” The reconstruction identifies the concern. It does not exempt the factual assertion from verification.

The trade-off is practical: a generous reconstruction can slow a fast exchange and may remove an easy rhetorical advantage. The benefit is that your eventual objection can land on a position the other person recognizes. The limit is ethical and epistemic: do not conceal material harms, invent evidence, or turn “I am worried” into “this policy is justified.”

Worked Trace: From Fidelity Check to Disagreement

Here is the Mina–Daniel exchange as a trace. Notice when the object changes.

1. Original claim — Mina: “A short visible trial can show whether participants accept a rule, but it may not show whether ordinary attention changes after the observation ends.”
2. Daniel’s first inference: “You reject experiments because you demand perfect data.”
3. Diagnosis: This is a strawman. It removes Mina’s distinction between acceptance and durability, then adds an absolute standard.
4. Daniel’s reconstruction: “You think the trial is useful for acceptance, but observation means it is weak evidence for durable attention. So we should not label the result a lasting improvement.”
5. Fidelity check — Mina: “Yes. I am not against the trial. I am against claiming it answered a different question.”
6. Daniel’s critique: “Then I disagree only about the next decision. A reversible trial may be worth continuing when acceptance is high, provided we measure attention later and report the limit honestly.”
7. Revision note: They now agree on the measurement limit and disagree about the action threshold. Neither has conceded the policy outcome.

This trace has a useful stopping rule: stop reconstructing once the speaker recognizes the conclusion, reasons, and material qualifier. Do not keep adding elegance until the claim becomes yours. Move to criticism or a question about the remaining decision.

Second check. After step 4, Mina replies: “Almost. I am less worried about the trial itself than about calling a temporary compliance effect ‘better attention.’” What should Daniel revise?

• Keep saying she rejects the trial, because the underlying concern is similar.
• Replace “weak evidence for durable attention” with the more specific concern about mislabeling compliance, then ask whether she accepts the revised summary.
• Upgrade her view into a demand for a six-month randomized study.

The middle choice is right. A correction is data. The goal is not to preserve Daniel’s earlier wording; it is to preserve Mina’s position more accurately.

Common Confusions

“If I steelman it, I have conceded it.” No. You have conceded only that you can state the target fairly. Daniel’s later objection shows how agreement on description can coexist with disagreement about action.

“The strongest version is whatever makes the argument hardest to refute.” Not necessarily. A view can be made harder to refute by adding claims, evidence, or moral commitments its speaker never offered. Strength without fidelity is substitution.

“Charity means overlooking harm so the conversation stays civil.” No. A faithful account preserves who bears the cost and what the policy does. Deleting those details is sanitization, not charity.

“A fidelity check is performative.” It can be, if you ask it after you have already decided what the other person believes. Used sincerely, it is a correction channel: “What did I miss?” is a way to update the working model.

Practice: A Four-Part Steelman

Use this claim from a neighborhood association discussion:

“The association should not choose every project by simple majority, because the same large block can approve projects that repeatedly bypass smaller streets.”

Write four short lines. Keep each line distinct.

1. Original claim: Copy it closely.
2. Faithful steelman: State the conclusion, strongest stated reason, and qualifier. Do not assume the speaker rejects all voting.
3. Fidelity question: Ask a sentence the speaker could correct: “Would you accept this as your view…?”
4. Your objection or question: Identify one target: evidence that the pattern is recurring, a proposed safeguard, or the trade-off between majority control and minority protection.

Use this rubric before you submit your work:

One possible model:

“You are not rejecting majority votes in general. You are worried that using only simple majority for every project can repeatedly exclude smaller streets, so you want some safeguard or rotation when the pattern persists. Would you accept that as your position?”

A precise critique might follow: “What evidence would show that this is a recurring pattern rather than a few regrettable cases, and which safeguard would protect minorities without giving any small group a permanent veto?” That question presses the policy design without pretending the speaker opposed voting.

In the next lesson, you will map the claim, reasons, and any hidden premises more explicitly. The steelman comes first because a map of the wrong argument is still wrong.

Resources

• [ARTICLE] Informal Logic — Stanford Encyclopedia of Philosophy — Focus: its account of charity and the “logically minimum” missing premise sets a boundary on reconstruction.
• [BOOK] Fights, Games, and Debates — Anatol Rapoport — Focus: read the book’s framework for conflict as a complement to practicing fair opposition.
• [ARTICLE] Thinking Rhetorically — Everyone’s an Author — Focus: use its listening-first orientation to connect reconstruction with rhetorical responsibility.

Key Takeaways

• A steelman is the strongest faithful version of a position, not the strongest nearby argument you can invent.
• Preserve the original, name conclusion/reasons/scope, label any minimum added link, and ask for a fidelity check before critique.
• Charity does not require agreement, false balance, or hiding harms; it makes the target of criticism more exact.
• After the speaker corrects your reconstruction, update it. Their correction is part of the evidence.