I’ve been pretty clear though that the fleshman isn’t about inventing ideas, it’s in fact meant to be more empirical than the steelman and something you have to go and observe the movement to extrapolate. The point I’m trying to convey is more about the “median” aspect. Some people are quite content to go on X/Twitter, find the Dumbest And Craziest Member of the outgroup, and point at it as a sign that the outgroup is Obviously Bad. This is technically “engaging with the outgroup”—the craziest member is still a real member! But it’s also a form of cherry picking evidence. Similar things can happen with the steelman—out of epistemic humility you might go and find the most sane, thoughtful representative of the idea you’re arguing against, someone so well articulated that you have a hard time finding fault in their version of the idea at all. That’s also engaging. But should you let that kind of thing discourage you from ever opposing that group again, on account of, they do have a point after all?
This is the thing I’m stressing. The opinion of the typical member is something you can only glean by observation, not pure thought, and is the one thing that’s going to matter the most. A classic example of this would be seen in many religious beliefs—the most educated theologians and scholars are not a good proxy of how it feels like to live in a society dominated by those beliefs. The typical believer is.
This is technically “engaging with the outgroup”—the craziest member is still a real member!
This is not centrally the case. You don’t need to go for the median to draw a sensible boundary and then retreat from it to the inside, with some margin. There will remain a lot of variation, on both content and quality, but you won’t risk the failure modes of non-centrality.
the most sane, thoughtful representative of the idea you’re arguing against, someone so well articulated
The criteria that grade arguments as sane/thoughtful/well-articulated should be suspect, because something might be such from your point of view, but not from inside the target worldview. If you optimize too strongly, you risk goodharting on your alien proxy measures of quality and pick a non-central representative as a “steelman” source.
so well articulated that you have a hard time finding fault in their version of the idea at all
Failing to find fault is very different from understanding an idea, digesting it well enough to keep developing it on your own. The purpose of steelmanning, in the (more useful, I think) distortionary variant of its meaning, is to build your own ideas regardless of whether they accurately represent and should be attributed to their original source/inspiration. This doesn’t work at all if you merely “have a hard time finding fault”.
This is not centrally the case. You don’t need to go for the median to draw a sensible boundary and then retreat from it to the inside, with some margin. There will remain a lot of variation, on both content and quality, but you won’t risk the failure modes of non-centrality.
I mean, I really have no objection to this—any kind of understanding of a whole distribution is always better than just understanding of individual simplifications of it (be average, median, or what have you). And it’s not like most of this stuff is literally something you can draw on a single 1D numerical axis, except for some very rare and specific policy positions such as “what tax rate in % should be applied to people above £100k” or such.
I was more gesturing at this “try to grasp what the common position is” concept. If you have a huge bimodal distribution where the movement is split almost 50⁄50 between two major interpretations of the belief then sure, both of those are more relevant than the rather meaningless mid-point. Apply pinch of salt and common sense as necessary.
The criteria that grade arguments as sane/thoughtful/well-articulated should be suspect, because something might be such from your point of view, but not from inside the target worldview.
Ultimately yes but there’s only so much relativism I can engage in. If someone posits a completely different worldview from mine down to “empirical knowledge isn’t valid”, what can I even do? Maybe it makes sense for them, but for me it’s insanity and as such disqualifying. This is different from “ok they share with me the basic principles of how the world works and how knowledge is accrued, and merely have different information, so if we engaged in enough mutual exchange we might come to an agreement or at least understanding”. Otherwise you’re not steelmanning, you’re just passively acknowledging any idea as just as good as any other, because for any idea there could be someone who thinks it’s perfectly coherent and sensible (for starters, usually, the person who voices it).
The purpose of steelmanning, in the (more useful, I think) distortionary variant of its meaning, is to build your own ideas regardless of whether they accurately represent and should be attributed to their original source/inspiration.
Steelmanning is used a lot of time as a way to engage with ideas/groups by giving maximum grace. Basically trying to play devil’s advocate to make sure that your own criticisms aren’t facile or purely the result of blind outgroup distrust, but grounded in genuine rational thinking. It’s not how you recommend to use it, but people do this, so I’m responding to that mainly.
I’ve been pretty clear though that the fleshman isn’t about inventing ideas, it’s in fact meant to be more empirical than the steelman and something you have to go and observe the movement to extrapolate. The point I’m trying to convey is more about the “median” aspect. Some people are quite content to go on X/Twitter, find the Dumbest And Craziest Member of the outgroup, and point at it as a sign that the outgroup is Obviously Bad. This is technically “engaging with the outgroup”—the craziest member is still a real member! But it’s also a form of cherry picking evidence. Similar things can happen with the steelman—out of epistemic humility you might go and find the most sane, thoughtful representative of the idea you’re arguing against, someone so well articulated that you have a hard time finding fault in their version of the idea at all. That’s also engaging. But should you let that kind of thing discourage you from ever opposing that group again, on account of, they do have a point after all?
This is the thing I’m stressing. The opinion of the typical member is something you can only glean by observation, not pure thought, and is the one thing that’s going to matter the most. A classic example of this would be seen in many religious beliefs—the most educated theologians and scholars are not a good proxy of how it feels like to live in a society dominated by those beliefs. The typical believer is.
This is not centrally the case. You don’t need to go for the median to draw a sensible boundary and then retreat from it to the inside, with some margin. There will remain a lot of variation, on both content and quality, but you won’t risk the failure modes of non-centrality.
The criteria that grade arguments as sane/thoughtful/well-articulated should be suspect, because something might be such from your point of view, but not from inside the target worldview. If you optimize too strongly, you risk goodharting on your alien proxy measures of quality and pick a non-central representative as a “steelman” source.
Failing to find fault is very different from understanding an idea, digesting it well enough to keep developing it on your own. The purpose of steelmanning, in the (more useful, I think) distortionary variant of its meaning, is to build your own ideas regardless of whether they accurately represent and should be attributed to their original source/inspiration. This doesn’t work at all if you merely “have a hard time finding fault”.
I mean, I really have no objection to this—any kind of understanding of a whole distribution is always better than just understanding of individual simplifications of it (be average, median, or what have you). And it’s not like most of this stuff is literally something you can draw on a single 1D numerical axis, except for some very rare and specific policy positions such as “what tax rate in % should be applied to people above £100k” or such.
I was more gesturing at this “try to grasp what the common position is” concept. If you have a huge bimodal distribution where the movement is split almost 50⁄50 between two major interpretations of the belief then sure, both of those are more relevant than the rather meaningless mid-point. Apply pinch of salt and common sense as necessary.
Ultimately yes but there’s only so much relativism I can engage in. If someone posits a completely different worldview from mine down to “empirical knowledge isn’t valid”, what can I even do? Maybe it makes sense for them, but for me it’s insanity and as such disqualifying. This is different from “ok they share with me the basic principles of how the world works and how knowledge is accrued, and merely have different information, so if we engaged in enough mutual exchange we might come to an agreement or at least understanding”. Otherwise you’re not steelmanning, you’re just passively acknowledging any idea as just as good as any other, because for any idea there could be someone who thinks it’s perfectly coherent and sensible (for starters, usually, the person who voices it).
Steelmanning is used a lot of time as a way to engage with ideas/groups by giving maximum grace. Basically trying to play devil’s advocate to make sure that your own criticisms aren’t facile or purely the result of blind outgroup distrust, but grounded in genuine rational thinking. It’s not how you recommend to use it, but people do this, so I’m responding to that mainly.