You are arguing against a version of the other side that does not exist
You are in a meeting, a debate, or a private deliberation inside your own head. You have a position. The other side has a position. You construct a version of their argument, identify its weaknesses, and dismantle it. You feel good. You won.
Except you did not engage their actual argument. You engaged a simplified, weakened version of it — one that was easy to defeat precisely because you built it to be defeated. Philosophers call this a straw man: a representation of an opposing view so thin and flimsy that knocking it down proves nothing. You did not test your belief against the strongest possible challenge. You tested it against the challenge you could already handle. And the difference between those two things is the difference between rigorous thinking and performance.
The practice of steel-manning — constructing the strongest possible version of an argument you disagree with — is the direct antidote. Before you attempt to resolve a contradiction between two beliefs, you make the best case for each side. Not a summary. Not a paraphrase. The version that the most thoughtful, most informed advocate of that position would recognize as their own — and possibly as better than what they could have articulated themselves.
This lesson makes a specific claim: you cannot resolve a contradiction you have not properly understood, and you cannot properly understand a position you have not genuinely tried to defend.
Rapoport's rules: the protocol for criticizing with integrity
The most precise formulation of this practice comes from game theorist Anatol Rapoport, popularized by philosopher Daniel Dennett in Intuition Pumps and Other Tools for Thinking (2013). Dennett presents four rules for engaging with an opposing position — rules that he calls "the best antidote for the tendency to caricature one's opponent":
- Re-express your target's position so clearly, vividly, and fairly that your target says, "Thanks, I wish I'd thought of putting it that way."
- List any points of agreement (especially if they are not matters of general or widespread agreement).
- Mention anything you have learned from your target.
- Only then are you permitted to say so much as a word of rebuttal or criticism.
Notice the sequencing. You do not earn the right to criticize until you have demonstrated that you understand the position well enough to improve its articulation, that you have identified common ground, and that you have acknowledged where the other position taught you something. Most people skip straight to step four. They begin with rebuttal and never pass through the three steps that would make their rebuttal meaningful.
Dennett explains why this protocol works: "One immediate effect of following these rules is that your targets will be a receptive audience for your criticism: you have already shown that you understand their positions as well as they do, and have demonstrated good judgment." But there is a second, deeper effect. The process of steel-manning often changes the steel-manner. When you genuinely try to express a position better than its holder can, you frequently discover that the position contains insights you had not noticed — that it is not just "not wrong" but partially right in ways that matter for your own thinking.
And there is a devastating third effect. As Dennett notes, "a heroic attempt to find a defensible interpretation of an author, if it comes up empty, can be even more devastating than an angry hatchet job." If you have given the opposing view every possible advantage and it still fails, your critique is far more powerful than if you had attacked a caricature.
Mill's argument: you do not understand your own position until you can argue the other
The deepest philosophical justification for steel-manning predates the term by more than 150 years. John Stuart Mill, in On Liberty (1859), made a case that remains the strongest argument for why you must argue both sides before resolving any contradiction:
He who knows only his own side of the case, knows little of that. His reasons may be good, and no one may have been able to refute them. But if he is equally unable to refute the reasons on the other side; if he does not so much as know what they are, he has no ground for preferring either opinion.
Mill goes further. It is not enough to hear the opposing arguments secondhand — filtered through your own teachers, presented alongside their refutations. "He must be able to hear them from persons who actually believe them... he must know them in their most plausible and persuasive form."
Mill's reasoning is epistemic, not social. This is not about being polite or fair-minded. It is about the structure of understanding itself. A belief you hold without understanding its strongest opposition is a belief you hold on faith, not on evidence. You may happen to be correct, but you cannot know that you are correct, because you have never subjected the belief to its most serious challenge. You are a boxer who has only sparred with partners who cannot hit.
This connects directly to the contradiction resolution work of Phase 19. When two of your beliefs conflict — when you hold "autonomous teams outperform managed ones" and "clear leadership direction is essential" — you cannot resolve the contradiction by arguing for your preferred side and dismissing the other. You must argue both sides at full strength. Only when each position is represented in its most defensible form can you see the actual shape of the contradiction and identify the variable that resolves it.
The principle of charity: interpretation as infrastructure
Steel-manning is the applied version of what philosophers call the principle of charity — a rule of interpretation with roots stretching back to ancient rhetoric but formalized in the twentieth century by Donald Davidson and Willard Van Orman Quine.
Davidson argued that charity is not merely a nice strategy for civil discourse. It is a precondition for understanding. If you interpret someone's statement in a way that makes it obviously foolish, you have almost certainly misunderstood what they meant. The principle of charity says: when an interpretation makes a speaker seem irrational and an alternative interpretation makes them seem rational, choose the rational interpretation. Not because people are always rational, but because assuming irrationality in others makes communication and genuine disagreement impossible.
The strongest version of this principle holds that charity is constitutive of meaning itself. You cannot even begin to interpret what someone is saying unless you assume, at the outset, that most of what they believe is true and most of what they say is rational. Without that assumption, you have no foothold from which to identify the specific claims where you disagree.
For contradiction resolution, this means: before you can productively engage a belief that conflicts with your own, you must interpret it charitably enough to see what makes it compelling. If you interpret it uncharitably — if you construct the weakest version — you will "resolve" the contradiction by defeating an opponent that does not exist. The real contradiction will remain, unexamined and unresolved, generating the cognitive costs that L-0372 will document.
"Consider the opposite": the experimental evidence
The practice of steel-manning has a precise parallel in cognitive science: the "consider the opposite" debiasing strategy, studied experimentally by Lord, Lepper, and Preston (1984).
In their landmark study, participants evaluated mixed evidence on a social issue — capital punishment. Without intervention, participants displayed biased assimilation: they accepted evidence supporting their existing position at face value and scrutinized evidence challenging it with hostile rigor. This is the default mode. Your brain applies asymmetric standards — gentle with confirming data, aggressive with disconfirming data — and calls the result "objective analysis."
Lord, Lepper, and Preston tested a simple intervention: ask participants to "consider the opposite." Before evaluating the evidence, subjects were instructed to ask themselves, "Would I have made the same evaluation if the study had produced results on the other side of the issue?" The results were striking. Participants who considered the opposite displayed significantly less biased assimilation than a control group — and the effect was stronger than simply instructing people to "be fair and unbiased." Telling people to be objective did not work. Forcing them to construct the opposing case did.
This is a critical finding for anyone attempting to resolve contradictions. General instructions to be fair-minded are ineffective because they do not change the underlying cognitive process. "Consider the opposite" works because it forces the specific cognitive operation that biased assimilation suppresses: generating arguments and evidence for the position you are inclined to reject. Steel-manning is the deliberate, disciplined practice of consider-the-opposite applied to your most important beliefs.
Mussweiler, Strack, and Pfeiffer (2000) extended this research to anchoring effects — the tendency for initial information to disproportionately influence subsequent judgments — and found the same result. Considering the opposite compensated for anchoring bias in ways that general debiasing instructions did not. The mechanism is consistent: when you are forced to construct a case for the alternative, you activate knowledge and evidence that your brain's default processing would have suppressed.
Red teaming: steel-manning as institutional practice
The military and intelligence communities formalized steel-manning decades before the rationalist community coined the term. They call it red teaming.
A red team is a group within an organization tasked with thinking like the adversary — arguing the opposition's case with full commitment, exploiting the vulnerabilities that the organization's own planners cannot see because they are trapped inside their own perspective. The U.S. Army's University of Foreign Military and Cultural Studies at Fort Leavenworth trains red team leaders explicitly in the skills of "alternative analysis" — generating interpretations and strategies that the primary planning team has not considered.
The purpose of a red team is not to be contrarian. It is to correct for groupthink — the tendency of cohesive groups to converge on a shared view and systematically suppress dissenting evidence. Irving Janis documented this phenomenon in his 1972 study of policy failures including the Bay of Pigs invasion, where Kennedy's advisors converged on a flawed plan because the group dynamics made it psychologically costly to voice opposition. After the failure, Kennedy institutionalized a form of red teaming in his decision-making process, assigning specific advisors to argue the opposing case in subsequent deliberations — including during the Cuban Missile Crisis, where the practice arguably prevented nuclear war.
The red teaming principle applies directly to individual contradiction resolution. When you hold two conflicting beliefs, your cognitive default is to argue for your preferred position and construct a weak version of the opposition. You are your own groupthink committee. Steel-manning is the practice of appointing your own internal red team — forcing yourself to argue the other side with the same rigor and commitment you would bring to your preferred position.
Debate pedagogy: what happens when you argue a position you disagree with
Structured debate programs have provided decades of evidence for what happens when people are forced to argue both sides of a question.
A critical feature of competitive debate formats is side assignment — debaters do not choose which side to argue. They are assigned a side, and they must argue it with full conviction regardless of their personal beliefs. This forces the exact cognitive operation that steel-manning requires: constructing the strongest possible case for a position you do not hold.
Research by Kuhn and Powell, testing a multi-year debate curriculum with middle school students, found that students who were repeatedly forced to argue both sides showed measurable improvements in critical thinking that persisted beyond the debate context. The mechanism is revealing: students forced to argue against their own position developed what Kuhn calls "meta-argumentation skill" — the ability to reason about arguments rather than merely within them. They learned to evaluate the structure of a case independently of whether they agreed with its conclusion.
A study with psychology students found an additional effect: students who were placed in groups arguing against their personal beliefs became less confident in their original positions — not because they were persuaded, but because they discovered that the opposing case had strengths they had never considered. This is the steel-manning effect in action. You do not necessarily change your mind. But your understanding of why the disagreement exists — what the actual crux of the contradiction is — deepens substantially.
Constitutional AI: machines that argue both sides of their own outputs
Anthropic's Constitutional AI framework provides a contemporary and perhaps surprising illustration of the steel-manning principle applied at scale.
In Constitutional AI, a language model is given a set of principles — a constitution — and tasked with evaluating its own outputs against those principles. Critically, the process involves the model generating both harmful and harmless responses to the same input, then using the constitutional principles to identify which response is better and why. The model is, in effect, steel-manning both sides of its own behavior: constructing the strongest case for a problematic response and the strongest case against it, then reasoning through the comparison.
The parallel to personal contradiction resolution is direct. When you encounter a contradiction between two of your beliefs, you face the same structural challenge that Constitutional AI addresses: you need to generate the best version of each side, hold them in comparison, and reason about which holds under what conditions — rather than defaulting to whichever side feels more comfortable or familiar.
Anthropic's red teaming research takes this further. In frontier AI safety evaluation, red teams are specifically trained to find the strongest possible attacks — the most compelling arguments for why the system might fail. A red team that only tests easy cases provides no useful safety signal. The value of the exercise is proportional to the quality of the opposing case. This is Mill's argument translated into machine learning practice: you do not know how robust your system is until you have subjected it to the strongest possible challenge.
The steel-manning protocol for contradiction resolution
Here is the practice that this lesson operationalizes. When you encounter a contradiction between two of your beliefs — the kind of productive tension that L-0361 taught you to recognize and that L-0362 through L-0370 taught you to classify — apply this protocol before attempting resolution:
Step 1: Write the strongest case for Side A. Not a summary. Not a bullet list. A full argument that the most sophisticated advocate of Side A would endorse. Include the best evidence. Identify the conditions under which this position is clearly correct. Anticipate the objections and address them. If you cannot produce a version that the other side would recognize, you do not yet understand what you are trying to resolve.
Step 2: Write the strongest case for Side B. Apply the same rigor to the opposing side. This is the harder step, because your cognitive defaults will resist it. Your brain will generate weak versions of the opposing case, steer you toward the arguments you already know how to rebut, and produce a sense of completion before you have actually done the work. Push past it. The quality of your resolution depends entirely on the quality of the weaker steel man.
Step 3: Read both cases as if you wrote neither. Physically set both arguments in front of you — on paper, on screen, side by side. Read them as if a colleague wrote them. Notice which one is actually stronger. Notice where the strongest argument on one side directly addresses the strongest argument on the other. Notice where they talk past each other — where Side A's best point and Side B's best point are answering different questions. That gap is often where the real contradiction lives.
Step 4: Identify the crux. The crux is the single factual or structural disagreement that, if resolved, would resolve the entire contradiction. After steel-manning both sides, the crux often becomes visible for the first time. Before steel-manning, you thought the disagreement was about X. After steel-manning, you realize it is actually about Y — a more specific, more tractable question that neither side had articulated clearly.
Step 5: Attempt resolution from the crux, not from your preferred side. With both positions at full strength and the crux identified, you can now reason about the contradiction without the bias that comes from having argued only one side. Your resolution — whether it is a synthesis, a context-dependent rule, or a genuine choice between the two — is grounded in the actual structure of the disagreement rather than in a fight between a strong argument and its weak caricature.
What steel-manning makes visible
The primary value of steel-manning is not persuasion and not politeness. It is epistemic. When you construct the strongest possible case for a position you oppose, three things happen that cannot happen any other way:
Hidden assumptions surface. Your original position contained assumptions you did not know were assumptions — they were invisible because the weakened version of the opposition never forced you to examine them. The steel man exposes them. You discover that your position depends on a specific factual claim you have never verified, or on a framing that excludes a legitimate concern, or on a definition that the other side is using differently.
The real disagreement sharpens. Most contradictions feel larger than they are because both sides are arguing against caricatures. Steel-manning compresses the disagreement to its actual size. You discover that you agree on far more than you thought, and that the remaining disagreement is specific, nameable, and investigable — not a vague sense that the other view is wrong.
Your confidence becomes calibrated. If you steel-man the opposing view and your position still holds, your confidence is earned. If the opposing view's steel man reveals weaknesses in your position, your confidence drops to where it should be. Either outcome is an improvement. Uncalibrated confidence — the kind produced by defeating straw men — is worse than being wrong, because it insulates you from the corrections that would eventually make you right.
The bridge to cost
You now have the tool that this phase has been building toward. L-0361 taught you that contradictions are data. L-0362 through L-0370 taught you to classify and disambiguate contradictions — to distinguish the surface tensions from the deep ones, to separate scope from level from time from perspective. This lesson gives you the practice that must precede any resolution attempt: argue both sides at their strongest before deciding anything.
The next lesson, L-0372, asks a harder question: what happens when you skip this step? What is the cost — cognitive, relational, epistemic — of living with contradictions that were never properly understood, let alone resolved? The answer is higher than most people expect. The dissonance does not stay contained. It leaks into decisions, relationships, and the structural integrity of your entire knowledge system.
Steel-manning is not a debate technique. It is a thinking discipline. And it is the last tool you need before confronting what unresolved contradictions actually cost.