Unstrawmanning Steelmanning

Contra Noah Smith on Steelmanning

Photo by Sakshi Adenwala on Unsplash

Strawmanning is when someone argues against a position by dismantling an artificially weak formulation of it. Steelmanning is the opposite of that: assigning particular attention to defeating the best case for an opposing view. Many people think that steelmanning is a noble endeavor that humans should aim towards, even if they regularly misfire. Noah Smith, however, has written an essay inveighing against steelmanning. The thrust of his argument is that those engaged in steelmanning confront a dilemma: it’s either ineffective or it’s misleading.

Impossibly Feeble Steel Men

Smith’s initial complaint about steelmanning is that steel men are invariably weak. Someone who disbelieves something isn’t ideally situated to outline why that view is so compelling. Moreover, a non-believer can’t muster the same logical ingenuity, desperate rationalization, and honed point-making that derives from honest convictions placed under threat.

Basically, a critic of an idea is inherently an unreliable guide to the best version of an idea. Personally, I don’t think mass deportation is a good policy. I’ve considered it. I won’t say it has zero appeal to me — law and order are important. I know it’s popular. But overall I just think it’s a bad idea . . . So if I tried to write an argument in favor of mass deportation, and if I kept it intellectually honest, I probably wouldn’t do a very good job of it.

But this is a paradox. How can things made of steel be so weak—would that really be steelmanning at all? Can you make a steel man out of rice paper? Well, no: this is the very distinction between strawmanning and steelmanning. It’s unfair to knock steel men by characterizing them as too wimpy and ascribe to them the failings of their straw-stuffed counterparts when their defining feature is precisely how they’re the opposite of that. It’s strawmanning steelmanning.

In fact, Smith’s heading for this section of his essay is “Steelmanning can easily turn into strawmanning.” But then it’s not really steelmanning, is it? And arguing that true steelmanning is especially tricky or rare doesn’t undermine claims that it’s laudable. I’m doubtful about the popular maxim that anything worth doing is difficult, but even a sluggish gadfly like me has to reject this mutant cri de coeur that something difficult obviously isn’t worth doing.

At best, Smith’s thesis that sincere attempts at steelmanning usually generate low-rent imitations and caricatures of steel men merely cautions against over-popularizing the technique (or maybe it cautions against trying to steel man something you cannot imagine putting compellingly). He doesn’t really land any blows against steelmanning per se. Maybe Smith’s correct that ventures in steelmanning are more liable to produce wheelspin and noise than to squarely engage with potent rebuttals, and therefore the practice can do more harm than good (the verschlimmbesserung theory of steelmanning, I guess). But if someone is so mentally inflexible and unappreciative of the temptations of salient counterarguments, and they’re discombobulated by the prospect of having to dismember a substantive counterpoint, then it’s unclear whether that person should be advocating for anything at all: it’s a shortcoming of that writer rather than a foible inherent to steelmanning.

Instead of imploring writers to construct surprisingly frail steel men, Smith recommends enlisting a bona fide adherent to champion the opposing perspective:

if you’re an ideologically neutral Washington Post editor whose only goal is to show public the best versions of each argument and then let the public decide, you probably don’t want Noah Smith writing in favor of mass deportation. You’ve hired the wrong man for the job.

He also points out that whatever he finds persuasive about opinions that he doesn’t endorse could diverge completely from whatever true believers find most convincing. In other words, one man’s steel man might be another’s straw man (and vice versa). But this is a feature of steelmanning rather than a bug. Probably, the people reading Noah Smith are more curious about and sympathetic towards whatever Noah Smith thinks is a forceful argument for something than about whatever tickles the instincts of the diehards.

So, if not for steelmanning, the most compelling rationales for maverick beliefs would be habitually de-emphasized, and the crucial task of stress-testing received wisdom would be outsourced to weirdo detractors. Whenever the rationale of one side is totally uninviting or inaccessible to their opponents is precisely when steelmanning is most warranted. Otherwise, groupthink rationales and confirmation bias would enjoy a monopoly on justification. To reject steelmanning here means hypothesizing that preaching to the choir is a better performing epistemic strategy than charitably framing opposing viewpoints.

I also wonder if Smith overestimates the heterogeneity of readers’ views or the heterogeneity of their intellectual diet, given how siloed and balkanized the info/opinion landscape is. For many, it’s possible that if they don’t tussle with a steel man, they may not encounter decent reasoning for anything they oppose at all. Smith complains later on about how steelmanning encourages sanewashing, but he gets this backwards, too: sanewashing is done by apologists covering up for fanatics, and if we cede the responsibility of justifying ideas to those with pre-existing loyalties to them, then sanewashing is the result.

Along with Smith’s worry that his steelmanning wouldn’t be convincing enough to court approval from his adversaries, he also thinks that catering to those adversaries would produce writing that’s too narrow-minded, given that “maximizing the number and diversity of the people reading your arguments is key to building a successful media business!” This is confused and nearly paradoxical. For writing to be so uniformly unappealing and unhelpful to all parties seems like it’d require a tailored bad-faith endeavor rather than just attempting to steel man something. Even if accidental, would such magnificent failure be the upshot of steelmanning or just clumsy and incapable writing? Beyond that, I dislike Smith’s recommendation that writers angle their takes to fish for the proper mix of readership. Is this why Smith writes what he does—to maximize the army of people who view him as reasonable? If so, it’s an ironic and strange admission in a piece that repeatedly exalts the primacy of intellectual honesty.

Misleading Steelmen

Smith also argues that steelmanning can be misleading in several ways: by repackaging noxious theories, papering over vulnerable assumptions, and promoting unrealistically ideal versions of policies. Unlike the weak-tea steelmanners above, misleading steelmanners function like mercenaries, recklessly espousing and promoting stuff they don’t even believe and optimizing for persuasion even at the price of resorting to subterfuge. This is the second horn of Smith’s steelmanning dilemma, and he differentiates between them by whether the steelmanning is intellectually honest:

I think we can see some downsides of two different types of steelmanning. Intellectually honest steelmanning is weak and unpersuasive, while intellectually dishonest steelmanning is likely to give terrible ideas too much respectability.

Smith objects to this guns-for-hire steelmanning where opinion writers role-play as attorneys for bad ideas. He isn’t the first to complain about sophists and pettifoggers—a tradition stretching back millennia—he’s merely recycling those criticisms and equipping steel man with a litigator’s briefcase. But even there Smith overplays his hand: lawyers shouldn’t obfuscate salient rejoinders or leave promising counterpoints unaddressed, and the legal system is geared toward uncovering truth anyway. In fact, Smith’s ultimate recommendations seems pretty close to being lawyerly. Relatedly, he objects to steelmanning insofar as it involves smart people lending brainpower to the proliferation of deleterious ideas, but this directly undermines his earlier thesis that steelmanning is perpetually ineffectual: how can the intelligentsia simultaneously be so inept at forcefully representing countervailing views that steelmanning is doomed to fail and yet be so skillful at it that it’s dangerous?

Furthermore, Smith claims that intellectually dishonest steelmanning requires adopting clay-footed viewpoints and obfuscating their critical weaknesses, but steelmanning requires marshaling the best counterarguments available—so that should exclude counterarguments with fatal defects. This critique relies on an incoherent notion that first-rate arguments include patently flawed premises, but if some argument is really just a castle built on unequivocally brittle foundations, then portraying it as your detractors’ crown jewel wouldn’t be steelmanning.

Per Smith’s reasoning, debaters who rely on the most implausible premises would be the most persuasive, since they can offload any portion of suspect logic onto their unstated assumptions and stash their flimsiness away, regardless of whether they’re wild/unconvincing. Can supporters of some controversial position just leave implausible assumptions unstated and unjustified without hampering their persuasiveness? If so, then their project is corrupted and their audience is too credulous—if anything, that’s an atmosphere that’s underexposed to steelmanning, not one that’s suffering from people being too charitable to their opponents. Leaving crucial and contestable premises unmentioned is neither rhetorically effective nor truth-conducive.¹

Smith’s other criticism of steelmanning is that it misleads people by defending idealized versions of policies even when that option is unrealistic and only loosely related to the proposals at hand (Smith uses the example of tariffs). But this conflates steelmanning a specific proposal with steelmanning a concept or policy in general. Steelmanning doesn’t require purposeful misrepresentation. If a particularized idea is on the table, then there’s a coherent version of steelmanning restricted to analyzing the merits of that proposal. Why must steelmanning entail some deceitful smoke-and-mirrors operation?

On What Steelmanning Is

The larger problem with Smith’s repudiation of intellectually dishonest steelmanning is that it’s an implausible and confused portrayal of what a steel man argument is. Strawmanning and steelmanning differ in the strength of their target, but both strategies are targeting something. Therefore, Steelmanning involves charitably framing the opposition by acknowledging and addressing high-quality counterarguments; it doesn’t require indefinitely feigning loyalty to the predilections of blockheads and know-nothings.

Thus, both horns of Smith’s dilemma present ironically straw-filled versions of steelmanning. I suppose the term “steelmanning” is up for grabs to some degree, but it’s essentially some kind of inversion of strawmanning. So, a complaint that steelmanning is pernicious because, secretly, it really is strawmanning, is just screwy and unconvincing. Moreover, to commit a straw man fallacy is to destroy an inferior version of an argument. To steel man something merely shifts the aim of your criticism to a more hardened target. Both tactics are employed en route to a takedown. Neither requires genuinely feigning belief in something. So Smith’s fears about jeopardizing intellectual honesty don’t check out.

Alternatively, complaining that steelmanning inadvertently leads to strawmanning (rather than fully equating the two ideas) still concedes that strawmanning is counterproductive. But then it’s worth identifying the downsides to strawmanning and thinking about whether evading those pitfalls doesn’t just wind up furthering the case for steelmanning. If blowing up an uncharitable cutout of your opposition is unconvincing or isn’t conducive to truth-finding, if it leaves room for facile rebuttals and gobbles up mental energy by forcing parties to constantly remind each other about their best points, and if you can appreciate this buffet of drawbacks, then doesn’t steelmanning seem like a virtuous aim or a fitting tincture?

Logically, you can either (1) articulate and engage with the premier counterarguments against your position (steelmanning), (2) represent the opposing viewpoint unfairly (strawmanning), or (3) totally ignore push back and counterpoints. The idea that people should leave the best arguments against their position unmentioned in the name of intellectual honesty isn’t defensible or even coherent, and Smith even admits that he tries not to ignore them, so presuming he makes judgments about which counterarguments to mention and selects them charitably, it’s arguable that this classifies as steelmanning. Is there some middle ground between strawmanning and steelmanning that acknowledges your opponents’ strongest counterarguments? If merely addressing the best counterpoints to your view—and therefore shoring up your own argument—qualifies as steelmanning, then Smith’s squeamishness about the practice is surely misplaced. By destroying the strongest form of the counterargument, you dispatch with its weaker formulations a fortiori.

He also suggests that steelmanning requires you to provide airtime to stupid, fringy, and noxious theories, but even supposing that steelmanning kooks is unwise, that hardly proves that you shouldn’t steelman at all. If something like identifying why the popular economic instincts motivating tariffs are misguided qualifies as steelmanning, then there’s no avoiding steelmanning if you ever intend to tackle specious arguments. Smith’s position imagines a sharp gulf between arguments for something and arguments against the alternative, but I’m unsure those two can be so neatly divided.

In sum, Smith is strawmanning steelmanning, and his essay could be improved with a little steelmanning of steelmanning, except that it would undermine the argument at hand. So, Smith’s essay inadvertently exemplifies the virtues of steelmanning by creating a dilemma of his own: by not steelmanning steelmanning, his essay suffers and is unconvincing, but doing otherwise would be self-undermining.

Conclusion

Does every crank’s screwball theorizing need to be steelmanned? Maybe not, especially if steelmanning requires a fulsome, systematic laying out of views you think are deranged. I’m not demanding that galaxy-brained op-ed columnists start LARPing as dummies and eccentrics and talking credulously about Jewish space lasers and hurricane machines or something: steelmanning doesn’t demand that. Sure, a myriad of outlandish views potentially fall short of deserving to be scrupulously disarmed, but if some issue is so epistemically toothless and uninviting, then the topic’s potentially not worth arguing over or writing about anyway.

How precisely we should demarcate heterodox theories that warrant recognition from those that can safely be ignored is a thorny question. Even being dismissive of wildly fallacious views can seemingly backfire and engender breakdowns in institutional trust, foment conspiratorial thinking, and beget a tidal wave of folks “doing their own research.” Giving up on even trying to steel man nonconformist viewpoints might exacerbate that situation, but either way, steelmanning doesn’t require anyone to permanently pretend as if they believe stuff that they disbelieve.

Smith’s takedown of steelmanning makes more sense if you suppose humans are otherwise evenhanded and dispassionate creatures who, without the undue influence of pro-steelmanning culture, would otherwise supply unbiased analysis, but a major goal in celebrating steelmanning is to correct the predominant habit of arguing against straw men. Maybe Smith is (or supposes himself to be) so phlegmatic and highly decoupling that this justification escapes his ken. But for bullheaded mortals, the reminder to occasionally unskew your evaluation of counterarguments is probably justifiable, and I see little risk of society overshooting and being too charitable in our disagreements.

He also claims that trusting the marketplace of ideas is more humble than representing both sides of an issue yourself:

I view this approach as a more humble alternative to steelmanning. When you steelman, you take it upon yourself to generate a little internal marketplace of ideas, where you scrimmage a synthetic version of your opponents’ arguments against your own. I prefer to leave things to the actual marketplace of ideas

But that little internal scrimmaging process he’s talking about is called “thinking,” and it’s something I consider to be important and good. This uncomfortable internal steelmanning process is arguably even more beneficial and necessary than the external steelmanning process analyzed above. To bypass any skepticism about your own reasoning is epistemically irresponsible. Smith complains that embracing steelmanning could mislead people, but refusing to internally steel man and interrogate your own beliefs will culminate in misleading yourself. By rejecting steelmanning to this degree, Smith is rejecting epistemic humility altogether.

Also, being an opinion writer automatically disqualifies you from being humble—it isn’t a humble pursuit, but the specialness of an opinion writer isn’t their views; it’s their reasoning and panache. The objective of the essayist isn’t to convey opinions magisterially. I read certain people because I’m interested in how their idiosyncratic brains digest and navigate the uncertain melange of reality. I have no interest in watching 200 IQ econ PhDs blowtorch pedestrian counterarguments into oblivion while they knowingly stash more persuasive steel men in their back pocket. Give me the fireworks! I’m not here for your authority; I’m here for your analysis.

Epistemic humility is more important for good professional opining than general humbleness, and to pontificate about your ideas without “scrimmag[ing] a synthetic version of your opponents’ arguments against your own” is just epistemic recklessness. And while Smith celebrates intellectual honesty, it’s even better to be both intellectually honest and correct. Sincerity has to be coupled with epistemic humility to function well, and inoculating your outlook from combat with a steel man puts that crucial modicum of humility at risk.

1. Hence my circling back on economic desert and free will recently, if you want an example of this.