(This comment is just notes that I took while I was reading, and not some particular single thought I put together after reflecting on the whole post.)
I’m so honored to be on your list of “unusually good rationalist communicators”. I really want to see your description of what each of us does 2-10x more than random LW users. Not just because I want you to talk about me; mostly I imagine it would be really educational for me to read some of these people’s writings with your perceptions in mind, especially if I first read an excerpt from each and try to write down for myself what they are doing unusually well. I certainly think my own writing is a lot stronger on some of your discourse norms than on others.
>Some ways you might feel when you’re about to break the Nth Guideline:
<3<3<3 that you included these
Question about Guideline 4: Where do you think my tendency (or Anna’s tendency, or Renshin’s tendency) to communicate in the form of interpretive poetry instead of unambiguous propositions falls with respect to Guideline 4? Or, more precisely: What thoughts do you have when you hold “interpretive poetry that results from attempts to express intuitions” up next to “make your claims clear”?
I think I have felt confused about this for years. Earlier today, while trying to share my cruxes after saying that I don’t want to cross post something to the EA forum, one of my cruxes was “I will be eaten by piranhas.” I’m not yet sure, in the sense of being able to communicate about it in a, um, standard way, what I mean by this, although I can absolutely belief report that it’s among my major concerns with posting there. It’s actually quite unusual for me to speak this way on LW, even though I happen to have done so today, but I think that outside of Lesswrong, I speak in “interpretive poetry” quite a lot. This is much of why I have by and large *avoided* participating in Lesswrong, preferring to share most of my writings as Google docs, emails, on Facebook, or as posts to my private blogs instead. I worry that my poetry is not welcome here, perhaps for very good reason; yet I also strongly suspect that there’s a ton of valuable information contained in my poetry, and it seems a shame not to share it here just because it often takes me several years after having a thought before I’m able to figure out what clear and unambiguous propositions correspond to the thought. I just seem to do most of my thinking at a level that is so far below words that translating all the way up into standard English is extremely difficult; should I indeed keep quiet, until I have done enough work that I can express myself “clearly”?
...Apparently I was so captivated by the rest of the essay that I stopped taking notes.
Something about how you have written this makes me feel like I’m accidentally training you to write differently, perhaps by making you review my essays and occasionally yelling at you* about yours. I dearly hope it is in fact for the better, and that I am not instead dragging you down. Anyway for better or worse, I found this piece delightful to read, and I expect to think about it and refer to it often in the future. [*Yes I am aware this is not a great description of my actual behavior, but it sort of feels to me like yelling compared to my baseline.]
Aw man, I was hoping that “Sabien’s Sins” was going to be “Here’s how I, Duncan Sabien, often fail at good rationalist discourse, and here are some actual examples of bad things I have said, and here is what it was like and how I think about it.”
>if you read further, please know that you are doing the equivalent of reading dictionary entries or encyclopedia entries and that the remaining words are not optimized for being Generically Entertaining To Consume.
Well, *I* found the entire thing highly entertaining, at any rate. Perhaps you ought to write encyclopedias.
This is probably my favorite essay you have written. Let’s get married at a trampoline park.
I would also love a more personalized/detailed description of how I made this list, and what I do poorly.
I think I have imposter syndrome here. My top guess is that I do actually have some skill in communication/discourse, but my identity/inside view really wants to reject this possibility. I think this is because I (correctly) think of myself as very bad at some of the subskills related to passing people’s ITTs.
Duncan has just replied on Facebook to my request for descriptions of what each person on his list does. He says it’s fine to copy his reply over here.
*
Julia Galef: something like, a science reporter whose hobby side project is doing science reporting for middle schoolers, and it’s in fact good and engaging and not stupid and boring. Wholesomeness, clarity, a tendency to correctly predict which parts of the explanation will break down for the audience and a corresponding slow and careful focus on those sections. Not unrelatedly: a sort of statesmanlike, reliable diplomacy; genuinely civilized debate; not the sort of person who will ever ever ever contribute to a discussion going off the rails. Grounding, stabilizing, sane-itizing.
Anna Salamon: something like how modern AI art programs can take a verbal prompt and spit out endless variations of image, and can take an image and spit out endless variations of interpretation and description. An ability to do co-Focusing, to find matches for felt senses, to find MISmatches between a felt sense and the preexisting model, and zero in on the delta, and rapidly put words to the delta, ad infinitum. An ability to make the proposition “some things just can’t be put into words” feel much less likely to be true.
Rob Bensinger: something like Data from Star Trek: The Next Generation. An ability to continuously produce relevant related information, an ability to maintain equanimity, an often-refreshing decoupling (in the decoupling vs. contextualizing sense). Someone who is unusually able to productively set aside human baggage and discuss things in depth, with mild manner, despite them perhaps being heavy or dangerous or deeply confusing.
Scott Garrabrant: something like an ability to externalize subtle and deep introspection, and to track the impacts of various pressures on his own cognition, and to hazard sound guesses as to the impacts of various pressures on the cognition of others. Fittingly: much closer to being a lens that can see itself than most people, and as a benefit, sometimes being able to serve as a mirror for other lenses to kind-of-sort-of see themselves.
Vaniver: something like an ability to synthesize realpolitik and the real world with the ivory tower; unusually capable of both recognizing *which* swathes of unfortunately-messy reality are relevant, and of keeping the conversation tethered to those swathes. “Would survive in the Game of Thrones universe without necessarily having to resort to evil” is kind of opaque, as a gesture toward why-Vaniver-is-a-good-communicator, but it fits the felt sense. Another way to say this is perhaps “comes to the table with the same sort of savvy that a veteran factory-floor engineer brings to a design discussion.”
Eliezer Yudkowsky: something like an ability to start from first principles and actually run the simulation forward, reliably and without skips and hallucinations and glossings-over, seeing what would happen rather than discovering that what you previously expected to happen of-course happens. Probably relatedly, the ability to *not* have his thinking visibly swayed by the hootings of the other monkeys; to neither cringe in fear nor seek applause.
Logan Strohl: something like the ability to cast “detect fuckery” and then also *say aloud* the results of the spell without getting torn to pieces by the other monkeys. A high sensitivity to things-being-askew coupled with an ability to express a) what the being-skewed feels like, in terms that allow others to recognize it, and b) hypotheses about what is causing the skewed-ness that are right or very close to right a shocking percentage of the time. Sort of like the love-child of the Anna-description and the Scott-description. Someone who speaks language as a second language, with both the drawbacks and the benefits that this provides (and it’s the benefits I’m pointing at, as valuable).
Oliver Habryka: something like courage and something like candor (I don’t mean to belittle; maybe it just IS those things but I can’t see inside his head so all I can speak to is what they resemble from the outside). Unwilling to let a falsehood pass, unafraid to voice a disagreement. The sort of person who, if he’s just “in the room,” I can trust that *I* have not accidentally said something false, because he will beep more reliably than almost anyone I know. Someone who actually flies the flag of truth, higher than any other flags (this is true of many of the people on this list but it’s conspicuously *representative* of Oliver’s particular virtue even moreso than the rest).
Kelsey Piper: something like the ability to bridge gaps, the ability to translate, the ability to host powwows, the ability to make-legible to one another *very* different views and cultures. Perhaps the best translator on the list. Able to extract the shape of the whole thought-tree from the mind of Person A, and to skillfully plant, water, grow, and trim the resulting thought-tree in the mind of Person B such that they are, in all the important ways, the *same* thought-tree, despite the two mental environments being very very different.
Nate Soares: something like the ability to track, and explicate, all of the uncertainties, and the uncertainties around those uncertainties, and the uncertainties around those as well; to make crisp and workable-with all of the possible flaws in a stack of conjecture. Not unrelatedly, the ability to track (and make trackable, for others, at least sometimes) all of the branches of dependency and possibility in a complex chain of thoughts; to note that A depends on B and C, and B depends on D, E, and F while C depends on D (also), G, H, and I, and that depending on the conclusion re: A something will *change* about J which informs both H and E.
Eric Rogstad: something like relentless good faith, the ability to be “on the side” of all parties in a meaningful and true sense, even if those parties are themselves opposed. Curious, and somehow *safe* in his curiosity; not someone whose curiosity comes with risk; not someone whose curiosity you will regret indulging (which is unfortunately not always the case!). In similar fashion: genuinely good-willed and good-natured; where someone like Julia or Kelsey might be low-conflict because Actually Diplomatic and someone like Scott or Logan might be low-conflict because Actually Has No Horse In This Race, Eric seems (to me) to embody and project something more like actual universal compassion, with correspondingly positive effects on discourse.
Spencer Greenberg: something like clean, something like straightforward, something like matter-of-fact. Stays firmly within the boundaries of his competence, keeps the ground under him, is cautious and reserved in his claims, stretching them only as far as observation will permit. Avoids trash fires and demon threads, possibly with a corresponding sacrifice in *which* topics he’s able to publicly discuss (mild contrast to my depiction of Rob above), but with the result that the Spencer-garden is well-kept and flourishing and tranquil and engaging. A Japanese tea garden of discourse.
Dan Keys: something like a Logan-esque detect fuckery, except instead of being focused on the human experience it’s focused on the interaction between information and summary. Reliably notes “I don’t think that Y conclusion is actually supported by X data.” In fact cares about the record, in fact cares about the evidence. An actual logician, an actual scientist. Trustworthy, by virtue of having filters that strike down 99 claims out of 100 and only allow the actually-deserving ones through.
should I indeed keep quiet, until I have done enough work that I can express myself “clearly”?
IMO, the answer here is a resounding “No!”
I think there’s a sort of unfortunate implication in the wording of the fourth guideline that I couldn’t quite erase without spending [so many words it ceased to be a simple statement].
But I do actually think “Do X or explicitly acknowledge that you can’t” means “Do X or Do Y” where Y is acknowledging that you can’t; I don’t actually think doing Y is worse than X, such that the fourth guideline says “Do X unless you suck.”
I mostly think of the fourth guideline as something like “everything in its place” or “everything with its proper epistemic status tags.”
I think there’s a T R E M E N D O U S amount of information that can be conveyed in poetry, that at least gets people looking in the right general direction or standing in the right general vicinity, and that a rationality community that taboo’d it because of its partial illegibility would be cutting off a major source of valuable intuition and wisdom.
(I especially think this because most of the skilled and generative original researchers I have met endorse thinking and speaking in poetry and would be horrified to find themselves in an environment where they could not.)
I think the Duty of an individual trying to not-undermine-rationality is to say “the following is poetry, because poetry is all I have; sorry; seems substantially better than nothing” at the start or the end of the poem. Then no one thinks the poetry is supposed to be airtight and fully legible, and thus the perceived standard of legibility is not undermined.
I think the short statement would be a lot weaker (and better IMO) if “inability” were replaced with “inability or unwillingness”. “Inability” is implying a hierarchy where falsifiable statements are better than the poetry, since the only reason why you would resort to poetry is if you are unable to turn it into falsifiable statements.
I think the Duty of an individual trying to not-undermine-rationality is to say “the following is poetry, because poetry is all I have; sorry; seems substantially better than nothing” at the start or the end of the poem
I just went to grab the link to Logan’s comment on the piranhas to note that in that context, I think including such a disclaimer would make the comment worse. I was sad to find that (I think?) they had edited it to have a disclaimer.
(there are other contexts where I think such a disclaimer is appropriate for logan-poetry-on-LW)
Huh, I guess I misremembered (glad I hedged there). If it was there originally I didn’t notice it which is perhaps evidence that it didn’t, in fact, make the comment noticeably worse.
Yeah, on second thought, please take that as the spirit of a recommendation and not the letter; the main threat vector I see is causing people confusion about what constitutes rigor or precision or a literal claim. I agree that there are a lot of cases where “this is not a literal claim” is pretty obvious on the surface to all-but-Lizardman-constant of the audience, and in those cases do not think a sign saying HERE COMES A POEM is always or even often indicated.
(This comment is just notes that I took while I was reading, and not some particular single thought I put together after reflecting on the whole post.)
I’m so honored to be on your list of “unusually good rationalist communicators”. I really want to see your description of what each of us does 2-10x more than random LW users. Not just because I want you to talk about me; mostly I imagine it would be really educational for me to read some of these people’s writings with your perceptions in mind, especially if I first read an excerpt from each and try to write down for myself what they are doing unusually well. I certainly think my own writing is a lot stronger on some of your discourse norms than on others.
>Some ways you might feel when you’re about to break the Nth Guideline:
<3<3<3 that you included these
Question about Guideline 4: Where do you think my tendency (or Anna’s tendency, or Renshin’s tendency) to communicate in the form of interpretive poetry instead of unambiguous propositions falls with respect to Guideline 4? Or, more precisely: What thoughts do you have when you hold “interpretive poetry that results from attempts to express intuitions” up next to “make your claims clear”?
I think I have felt confused about this for years. Earlier today, while trying to share my cruxes after saying that I don’t want to cross post something to the EA forum, one of my cruxes was “I will be eaten by piranhas.” I’m not yet sure, in the sense of being able to communicate about it in a, um, standard way, what I mean by this, although I can absolutely belief report that it’s among my major concerns with posting there. It’s actually quite unusual for me to speak this way on LW, even though I happen to have done so today, but I think that outside of Lesswrong, I speak in “interpretive poetry” quite a lot. This is much of why I have by and large *avoided* participating in Lesswrong, preferring to share most of my writings as Google docs, emails, on Facebook, or as posts to my private blogs instead. I worry that my poetry is not welcome here, perhaps for very good reason; yet I also strongly suspect that there’s a ton of valuable information contained in my poetry, and it seems a shame not to share it here just because it often takes me several years after having a thought before I’m able to figure out what clear and unambiguous propositions correspond to the thought. I just seem to do most of my thinking at a level that is so far below words that translating all the way up into standard English is extremely difficult; should I indeed keep quiet, until I have done enough work that I can express myself “clearly”?
...Apparently I was so captivated by the rest of the essay that I stopped taking notes.
Something about how you have written this makes me feel like I’m accidentally training you to write differently, perhaps by making you review my essays and occasionally yelling at you* about yours. I dearly hope it is in fact for the better, and that I am not instead dragging you down. Anyway for better or worse, I found this piece delightful to read, and I expect to think about it and refer to it often in the future. [*Yes I am aware this is not a great description of my actual behavior, but it sort of feels to me like yelling compared to my baseline.]
Aw man, I was hoping that “Sabien’s Sins” was going to be “Here’s how I, Duncan Sabien, often fail at good rationalist discourse, and here are some actual examples of bad things I have said, and here is what it was like and how I think about it.”
>if you read further, please know that you are doing the equivalent of reading dictionary entries or encyclopedia entries and that the remaining words are not optimized for being Generically Entertaining To Consume.
Well, *I* found the entire thing highly entertaining, at any rate. Perhaps you ought to write encyclopedias.
This is probably my favorite essay you have written. Let’s get married at a trampoline park.
I would also love a more personalized/detailed description of how I made this list, and what I do poorly.
I think I have imposter syndrome here. My top guess is that I do actually have some skill in communication/discourse, but my identity/inside view really wants to reject this possibility. I think this is because I (correctly) think of myself as very bad at some of the subskills related to passing people’s ITTs.
Duncan has just replied on Facebook to my request for descriptions of what each person on his list does. He says it’s fine to copy his reply over here.
*
Julia Galef: something like, a science reporter whose hobby side project is doing science reporting for middle schoolers, and it’s in fact good and engaging and not stupid and boring. Wholesomeness, clarity, a tendency to correctly predict which parts of the explanation will break down for the audience and a corresponding slow and careful focus on those sections. Not unrelatedly: a sort of statesmanlike, reliable diplomacy; genuinely civilized debate; not the sort of person who will ever ever ever contribute to a discussion going off the rails. Grounding, stabilizing, sane-itizing.
Anna Salamon: something like how modern AI art programs can take a verbal prompt and spit out endless variations of image, and can take an image and spit out endless variations of interpretation and description. An ability to do co-Focusing, to find matches for felt senses, to find MISmatches between a felt sense and the preexisting model, and zero in on the delta, and rapidly put words to the delta, ad infinitum. An ability to make the proposition “some things just can’t be put into words” feel much less likely to be true.
Rob Bensinger: something like Data from Star Trek: The Next Generation. An ability to continuously produce relevant related information, an ability to maintain equanimity, an often-refreshing decoupling (in the decoupling vs. contextualizing sense). Someone who is unusually able to productively set aside human baggage and discuss things in depth, with mild manner, despite them perhaps being heavy or dangerous or deeply confusing.
Scott Garrabrant: something like an ability to externalize subtle and deep introspection, and to track the impacts of various pressures on his own cognition, and to hazard sound guesses as to the impacts of various pressures on the cognition of others. Fittingly: much closer to being a lens that can see itself than most people, and as a benefit, sometimes being able to serve as a mirror for other lenses to kind-of-sort-of see themselves.
Vaniver: something like an ability to synthesize realpolitik and the real world with the ivory tower; unusually capable of both recognizing *which* swathes of unfortunately-messy reality are relevant, and of keeping the conversation tethered to those swathes. “Would survive in the Game of Thrones universe without necessarily having to resort to evil” is kind of opaque, as a gesture toward why-Vaniver-is-a-good-communicator, but it fits the felt sense. Another way to say this is perhaps “comes to the table with the same sort of savvy that a veteran factory-floor engineer brings to a design discussion.”
Eliezer Yudkowsky: something like an ability to start from first principles and actually run the simulation forward, reliably and without skips and hallucinations and glossings-over, seeing what would happen rather than discovering that what you previously expected to happen of-course happens. Probably relatedly, the ability to *not* have his thinking visibly swayed by the hootings of the other monkeys; to neither cringe in fear nor seek applause.
Logan Strohl: something like the ability to cast “detect fuckery” and then also *say aloud* the results of the spell without getting torn to pieces by the other monkeys. A high sensitivity to things-being-askew coupled with an ability to express a) what the being-skewed feels like, in terms that allow others to recognize it, and b) hypotheses about what is causing the skewed-ness that are right or very close to right a shocking percentage of the time. Sort of like the love-child of the Anna-description and the Scott-description. Someone who speaks language as a second language, with both the drawbacks and the benefits that this provides (and it’s the benefits I’m pointing at, as valuable).
Oliver Habryka: something like courage and something like candor (I don’t mean to belittle; maybe it just IS those things but I can’t see inside his head so all I can speak to is what they resemble from the outside). Unwilling to let a falsehood pass, unafraid to voice a disagreement. The sort of person who, if he’s just “in the room,” I can trust that *I* have not accidentally said something false, because he will beep more reliably than almost anyone I know. Someone who actually flies the flag of truth, higher than any other flags (this is true of many of the people on this list but it’s conspicuously *representative* of Oliver’s particular virtue even moreso than the rest).
Kelsey Piper: something like the ability to bridge gaps, the ability to translate, the ability to host powwows, the ability to make-legible to one another *very* different views and cultures. Perhaps the best translator on the list. Able to extract the shape of the whole thought-tree from the mind of Person A, and to skillfully plant, water, grow, and trim the resulting thought-tree in the mind of Person B such that they are, in all the important ways, the *same* thought-tree, despite the two mental environments being very very different.
Nate Soares: something like the ability to track, and explicate, all of the uncertainties, and the uncertainties around those uncertainties, and the uncertainties around those as well; to make crisp and workable-with all of the possible flaws in a stack of conjecture. Not unrelatedly, the ability to track (and make trackable, for others, at least sometimes) all of the branches of dependency and possibility in a complex chain of thoughts; to note that A depends on B and C, and B depends on D, E, and F while C depends on D (also), G, H, and I, and that depending on the conclusion re: A something will *change* about J which informs both H and E.
Eric Rogstad: something like relentless good faith, the ability to be “on the side” of all parties in a meaningful and true sense, even if those parties are themselves opposed. Curious, and somehow *safe* in his curiosity; not someone whose curiosity comes with risk; not someone whose curiosity you will regret indulging (which is unfortunately not always the case!). In similar fashion: genuinely good-willed and good-natured; where someone like Julia or Kelsey might be low-conflict because Actually Diplomatic and someone like Scott or Logan might be low-conflict because Actually Has No Horse In This Race, Eric seems (to me) to embody and project something more like actual universal compassion, with correspondingly positive effects on discourse.
Spencer Greenberg: something like clean, something like straightforward, something like matter-of-fact. Stays firmly within the boundaries of his competence, keeps the ground under him, is cautious and reserved in his claims, stretching them only as far as observation will permit. Avoids trash fires and demon threads, possibly with a corresponding sacrifice in *which* topics he’s able to publicly discuss (mild contrast to my depiction of Rob above), but with the result that the Spencer-garden is well-kept and flourishing and tranquil and engaging. A Japanese tea garden of discourse.
Dan Keys: something like a Logan-esque detect fuckery, except instead of being focused on the human experience it’s focused on the interaction between information and summary. Reliably notes “I don’t think that Y conclusion is actually supported by X data.” In fact cares about the record, in fact cares about the evidence. An actual logician, an actual scientist. Trustworthy, by virtue of having filters that strike down 99 claims out of 100 and only allow the actually-deserving ones through.
IMO, the answer here is a resounding “No!”
I think there’s a sort of unfortunate implication in the wording of the fourth guideline that I couldn’t quite erase without spending [so many words it ceased to be a simple statement].
But I do actually think “Do X or explicitly acknowledge that you can’t” means “Do X or Do Y” where Y is acknowledging that you can’t; I don’t actually think doing Y is worse than X, such that the fourth guideline says “Do X unless you suck.”
I mostly think of the fourth guideline as something like “everything in its place” or “everything with its proper epistemic status tags.”
I think there’s a T R E M E N D O U S amount of information that can be conveyed in poetry, that at least gets people looking in the right general direction or standing in the right general vicinity, and that a rationality community that taboo’d it because of its partial illegibility would be cutting off a major source of valuable intuition and wisdom.
(I especially think this because most of the skilled and generative original researchers I have met endorse thinking and speaking in poetry and would be horrified to find themselves in an environment where they could not.)
I think the Duty of an individual trying to not-undermine-rationality is to say “the following is poetry, because poetry is all I have; sorry; seems substantially better than nothing” at the start or the end of the poem. Then no one thinks the poetry is supposed to be airtight and fully legible, and thus the perceived standard of legibility is not undermined.
I think the short statement would be a lot weaker (and better IMO) if “inability” were replaced with “inability or unwillingness”. “Inability” is implying a hierarchy where falsifiable statements are better than the poetry, since the only reason why you would resort to poetry is if you are unable to turn it into falsifiable statements.
I changed it to say “aren’t doing so (or can’t).”
I just went to grab the link to Logan’s comment on the piranhas to note that in that context, I think including such a disclaimer would make the comment worse. I was sad to find that (I think?) they had edited it to have a disclaimer.
(there are other contexts where I think such a disclaimer is appropriate for logan-poetry-on-LW)
>I was sad to find that (I think?) they had edited it to have a disclaimer.
(Actually it originally had that disclaimer, or else I probably wouldn’t have posted it.)
Huh, I guess I misremembered (glad I hedged there). If it was there originally I didn’t notice it which is perhaps evidence that it didn’t, in fact, make the comment noticeably worse.
Yeah, on second thought, please take that as the spirit of a recommendation and not the letter; the main threat vector I see is causing people confusion about what constitutes rigor or precision or a literal claim. I agree that there are a lot of cases where “this is not a literal claim” is pretty obvious on the surface to all-but-Lizardman-constant of the audience, and in those cases do not think a sign saying HERE COMES A POEM is always or even often indicated.