Should you steelman people you disagree with?

Short of time? Read the key takeaways.

🛠️ Steelmanning means developing the strongest, most defensible version of a view you oppose. This can involve improving an argument beyond what its original advocate had in mind.

🔍 Engaging with strong opposing arguments helps you test your own beliefs and get to the truth. Understanding how well rival views explain the evidence puts you in a better position to judge whether your own view is justified.

🤝 Steelmanning can help people develop their ideas together. Done respectfully, it lets you contribute expertise, uncover nuances, and turn tentative suggestions into more robust arguments. This happens far more often than you might think.

🗣️ Check your understanding before trying to improve someone’s argument. Imagining another person’s perspective can increase confidence without improving accuracy, especially when their worldview differs fundamentally from yours. Whenever possible, ask them directly and check that you’ve understood.

💬 An attempted improvement can come across as an insult. Offering someone a “better” argument can imply that their own is inadequate. Productive discussion sometimes calls for focusing on understanding their view accurately.

⏳ Choose carefully which arguments deserve your attention. Steelmanning can waste time, encourage poorly developed contributions, or reward people arguing in bad faith. Sometimes it’s better to move on.

🎯 A useful steelman can reveal the crux of a disagreement and the evidence needed to address it. The article addresses the “crocoduck” argument against evolution by (a) turning a caricature of evolution into a substantive question about transitional fossils and (b) showing how discoveries such as Tiktaalik help answer it.

When the actor Kirk Cameron sought to disprove evolution during a debate on ABC, he argued that:

“Science has never found a genuine transitional form, that is, one kind of animal crossing over into another kind, either living or in the fossil record. And there are supposed to be billions of them. Now what I’m about to show you does not exist. These were actually created by our graphic artists. But I want you to keep your eye out for this because this is what evolutionists have been searching for for hundreds of years. All right? And if you find one of these you could become rich and famous. So here’s some transitional forms. This is called the crocoduck.”

He then showed the audience this image:

Source: ABC Nightline debate

Unfortunately, this is a misrepresentation of what the theory of evolution claims. In order for the absence of crocodile-to-duck transitional forms to count against evolution, it would have to be the case that the theory of evolution (or our best understanding of its consequences) claims that crocodiles evolved into ducks or ducks into crocodiles. But that is not what it claims. 

And, indeed, many transitional forms that actually are predicted by evolution have been discovered (for example, some are discussed here and here). One such form was a transitional fossil between dinosaurs and birds, called “Archæopteryx”, that was discovered just two years after Darwin’s On the Origin of Species was published. 

Now, here are two questions for you:

1. Can you extract a better argument out of what Cameron said?

2. Why would you want to do that?

The argument against evolution offered by Kirk Cameron was a “straw man”; it presented a weakened and inaccurate caricature of an opposing view (evolution), flimsy enough to easily topple over. Cameron could rely on the audience members finding the idea of crocoducks roaming about in the wild quite ridiculous, so by caricaturing evolution as committed to the existence of such creatures, his argument appears aimed at provoking the audience to dismiss evolution as similarly ridiculous.

The straw man is a widely understood concept. You’ve probably heard of it before. But what’s much less well-known (and commonly misunderstood) is its opposite: the steel man. There’s no single universal definition of “steelmanning”, but it’s usually defined along the lines of:

In principle, you can do this in any situation where you’re engaging with views you disagree with – whether that’s in person, in internet comments, in an essay, etc.

In this article, we’re going to examine the important practice of “steelmanning” arguments, which turns out to be a lot more difficult and dangerous than you might imagine. We’re going to tell you about:

1. Why you should steelman

2. What traps to avoid when steelmanning

Then we’ll put those insights into action by giving and evaluating:

3. A steelman version of Cameron’s argument against evolution

Why should you steelman views you disagree with?

1. Steelmanning helps you find the truth

The classic argument for steelmanning comes from the philosopher John Stuart Mill, who argued that if you don’t know the strongest cases in favor of the views you oppose, then you don’t know how well those views can account for the same evidence that you’re relying on to justify your own view. If you only evaluate weakened, straw versions of opposing views, then the evidence will look harder to explain without your view than it really is, and you won’t really know the strength of the evidence for your own position. 

A related argument for steelmanning is this: When you’re trying to determine whether something is true or not, stronger arguments give you much more help than weak ones. For example, imagine you’re a defense lawyer, and you’re trying to prove in court that your client didn’t murder someone. Weak arguments like this one won’t be much help to the jurors:

This is weak because the premise being true barely makes the conclusion more likely to be true. There are so many ways the premise could be true and the conclusion false (very often when a person murders someone, there is no record of them threatening that person). On the other hand, stronger arguments like this one tell you much more:

In this case, as long as the premise is true, the conclusion is close to definitively entailed. There are still ways it could be false (e.g., the murder method involved a delay long enough for your client to leave the country), but this stronger argument gives the jurors much more help figuring out the truth of the matter.

Steelmanning helps you get to the truth because it helps you find the strongest relevant arguments, and it's typically only the strongest arguments that really matter when trying to determine the truth.

2. Steelmanning helps develop ideas

When you steelman the views of others (as long as you do it in a respectful and non-dismissive way) you can help them improve their own ideas.

When I’m debating topics in history or international relations or physics with my very smart friends who research and teach those topics, I can say what I’m thinking, but I benefit from those friends helping me refine my thoughts. They can add depth and show me nuances that I wouldn’t have spotted myself. They can help me turn tentative ideas into something more robust. When they do so, they’re steelmanning what I’m saying. They’re helping me develop my ideas. And I do the same for them in return when we debate philosophy (my field of study).

This doesn’t only happen in conversations with friends: steelmanning as a mechanism for developing someone else’s ideas can happen any time there is a difference in relevant expertise between two people debating a topic in good faith and a collaborative and respectful manner. For example, teachers do this kind of steelmanning of students’ ideas and questions all the time. Engineers and designers do it with clients’ proposals. Editors do it with authors’ works. Experienced employees do it with new employees’ suggestions. And so on.

What should you avoid when steelmanning?

1. Avoid False Confidence

David Lewis and David Armstrong were two of the most highly regarded analytic philosophers of the last 100 years. They were rivals and friends; despite disagreeing on many things, they also helped each other out a lot. For example, one of Lewis’s papers began with the declaration:

“I think there is more to be said for Armstrong’s theory than he himself has said.”

In that paper, Lewis showed that Armstrong’s theory could be used to solve various problems in philosophy, in ways nobody had yet realized. Lewis still did not adopt Armstrong’s theory, but that paper did much to strengthen it.

It’s nice to imagine that we are all like Lewis; capable of steelmanning our opponent’s views so cogently that our insights become a key part of their theories. In practice, however, empirical evidence suggests reasons to think people are not as good at this as they want to be.

Researchers distinguish between:

• Perspective taking: imagining the point of view of someone else

• Perspective getting: seeking the point of view of someone else directly in conversation

There is evidence that perspective taking helps reduce all sorts of biases, such as hindsight bias and overconfidence bias. However, perspective taking does not appear to be associated with improvements in accuracy. Perspective getting does. The scary thing is that perspective taking can feel like it improves the accuracy of your understanding of others’ views without actually doing so. Of course, it’s still often worthwhile to try to imagine other people's perspectives, but it’s also worthwhile to acknowledge that doing so can be a glorified form of guessing, and is typically inferior to actually getting those perspectives from the source.

People in the rationalist community have noticed this for a while. The blogger Ozy Brennan wrote an article “Against Steelmanning” in which he warned about cases involving “fundamental worldview differences”. These are cases where the person you’re disagreeing with has a worldview so fundamentally different from yours that you are not able to adopt their perspective accurately enough to steelman it. In such cases, what you might think is steelmanning is really just creating a distorted version of their argument that is most persuasive to people with your worldview, without really grasping the root of their beliefs. Here’s an adaptation of one of Brennan’s examples:

All of this serves as a warning: If you’re attempting to steelman someone’s view, be wary of false confidence. You might think that you’ve managed to take their perspective and improve it, but there’s always a danger that you’ve strawmanned them by accident. In principle, it might be possible to steelman someone’s views without understanding them, but understanding the view is likely to help you do a better job. 

It might help to ask yourself whether you would be able to pass what’s known as an “ideological Turing test”, before attempting to improve their view. That is, would you be able to articulate their view in such a way that your version cannot be discerned from theirs, by people who share their perspective? That’s a high bar.

If it’s possible, check with your interlocutor that you really do understand where they’re coming from. If you are mischaracterizing a view because you can’t accurately imagine it, you’re not going to be able to find the best version of it.

2. Avoid being insulting

Another good point on this topic coming from the rationalist community is that steelmanning can be interpreted as rude or combative. By doing it, you’re essentially saying “here’s a better argument than the one you were making”, which is (at least implicitly) putting down the argument you’re engaging with. As such, it’s not always going to be appropriate in contexts where you’re directly talking to an advocate of the view you oppose.

If you want to have a productive debate with someone whom you suspect is likely to interpret your actions as hostile, then you’ll need to take extra care not to be insulting. Focusing on the most robust steel man might be less important than focusing on the most accurate version of their view, depending on the context. That’s a judgment you’ll have to make on a case-by-case basis.

3. Avoid Wasting Time

Aikin and Casey point out that, since steelmanning is the opposite of strawmanning, we might expect it to produce opposite problems. One problem with strawmanning is that it results in us ignoring arguments that are worth our time to consider. When Kirk Cameron caricatured evolution by talking about crocoducks, he missed out on the chance to engage with worthwhile arguments about the genuine explanatory and predictive power of the theory. The opposite of that problem is this: steelmanning can result in us paying attention to perspectives that are not worth our time. 

Imagine a lecturer who always steelmans the points made by their students, regardless of quality. When a student offers a particularly dubious contribution, steelmanning runs the risk of wasting class time, encouraging futile lines of thought, or even robbing the student of the opportunity to refine their view themselves. Of course, the student ought not to be cruelly dismissed, but there are times when it is prudent not to dwell on a poorly thought-through or unhelpful perspective. 

And when the people you’re arguing with are doing so in bad faith, spending time and energy on finding the best versions of their points can be an even greater waste. If you encounter an internet troll whose goal is to create outrage or be a nuisance, you're better off simply ignoring them rather than engaging with their perspective.

What’s the steelman version of the crocoduck? 

Now instead of looking for literal crocoducks (as Kirk Cameron suggested), let’s finish up by looking for a steelman version of the crocoduck argument against evolution. A steel crocoduck, if you will (we might also reserve this phrase for any time we steelman a strawman). We can then attempt to evaluate how well evolutionary theory handles the strongest version of such an argument. We hope this exercise will demonstrate how to apply some of the insights above.

A "steel crocoduck" - which is what you get

when you steelman a straw man.

Since we want to avoid false confidence, it will help to make sure we understand Cameron’s view as thoroughly as possible. We’re not able to check directly with him, but by looking into other things he has said or published on this topic, we can gain a deeper understanding. For instance, in this later YouTube video, we find the claim:

“Now, of course, evolution doesn’t claim that the crocoduck ever existed, but if evolutionists can make up imaginary missing links, why can’t we?”

This casts the crocoduck argument in a new light. It seems that we are to take it as parody, rather than a literal representation of what Cameron believes is claimed by the theory of evolution.

In that later YouTube video, we also see more detail added to their argument. Cameron and his collaborators make the claim that no genuine transitional fossils have been found, and they point to cases of alleged transitional fossils that really did turn out to be fakes. For example, this article discusses how Archaeoraptor (not to be confused with Archaeopteryx, which was mentioned near the start of this article) turned out to be a forgery built out of the bones of two different specimens.

Again, it might be tempting to interpret this argument as very weak. You might think they’re simply ignorant of all the genuine transitional fossils that have been discovered – or that they are deliberately cherry picking only the fraudulent handful and ignoring the huge number of genuine ones. But since we want to steelman their argument, let’s assume they’re not ignorant or cherry picking. Instead, we can assume they think the fraudulent cases erode confidence in other fossils and that they find other fossils unconvincing (such as is said by Cameron’s collaborator, Ray Comfort, in blog posts like this one).

There are usually going to be multiple ways you could steelman an argument, and this is no exception. For the sake of brevity, here’s just one potential steelman of the crocoduck argument:

The crux of this argument is whether or not the evidence adequately supports the authenticity and interpretation of fossils that scientists claim are transitional.

This argument suggests a test: If the theory of evolution is actually true, scientists should be able to use their knowledge of the fossil record to predict the existence of transitional species that haven’t yet been found. They should then be able to go out and actually discover fossils of those species at least some of the time.

Passing this test would be very strong evidence for the transitional interpretation of those fossils (and therefore very strong evidence of evolution). It might also be worrying for the theory if this never happened – though we should make sure our expectations account for the fact that fossilization is quite rare because it requires very specific conditions.

This is actually a fairly strong argument and a cogent test that we've developed from a straw man! So, how does evolution fare against it?

Well, as of the early 1990s, the fossil record left an important gap between fish and land-walking “tetrapods” (four-limbed vertebrates). This gap existed despite the expectation (based on evolutionary theory plus observations from geology and other fossils) that tetrapods evolved from lobe-finned fish. 

In other words, a transitional species was expected, but its matching fossil had not yet been found.

Image adapted from Figure 7 in this article.

Neil Shubin, Edward Daeschler, Farish Jenkins Jr., and their colleagues began Arctic fieldwork in 1999 to search for fossils illuminating this transition. It made sense to look specifically at rocks from the Devonian period because the Devonian period is full of fish (indeed, it’s poetically referred to as “the age of fishes”) and tetrapod fossils are found in rocks from late in that period but not in rocks from early in that period. Geological maps helped them identify promising Devonian sedimentary rocks, and, after refining their search, their team discovered Tiktaalik roseae in 2004, right where they expected it to be. Shubin recounts the fascinating story in this video.

Tiktaalik displayed the broadly expected combination of fish-like and tetrapod-like features. For example, like a fish, it had scales, fins, and gills. But like a tetrapod, it had a flattened head with upward-positioned eyes and a neck.

Cast of the Tiktaalik holotype in the Field Museum, Chicago - image from Wikipedia.

The case of Tiktaalik demonstrates the theory of evolution making clear predictions about transitional fossils, and those predictions later being fulfilled. The fact that many more Tiktaalik fossil specimens have been discovered since (more than 60 specimens, by the year 2020) and the specimens are housed in museums where lots of independent researchers can (and do!) go to examine them supports being confident in their authenticity. This, and other cases like Tiktaalik, show that the theory of evolution passes the test posed by the steel crocoduck argument.

Image adapted from Figure 7 in this article.

Hopefully, this exercise in steelmanning has shown some of the methods and benefits of steelmanning in action. By steelmanning Cameron’s crocoduck argument we:

• bypassed superficial arguments about obviously false claims like “evolution predicts the crocoduck.”

• located a deeper, potential crux of the disagreement more precisely and identified the evidence needed to address it.

• were required to wrestle with deeper concerns that people have about the evidence for evolution and to make a much more robust case in its favor.

This kind of practice helps us deepen our understanding of our own views and others’, and it helps us more reliably get to the truth.

If you enjoyed this article, you might enjoy our “Can You Detect Weak Arguments?” quiz. Find out how well you can spot misleading rhetoric and recognize your own bias using real-world examples:

Launch “Can You Detect Weak Arguments?”