The following is an interview I had informally with Duncan Sabien back in early 2020 when I was trying to improve my writing. I thought Duncan gave some great, actionable tips, so I’m sharing it here for other people trying to improve their writing. 

The transcript has been lightly edited for clarity and readability, and is crossposted on my blog. 


Lynette Bye: I’m curious, when did you get started writing? 

Duncan Sabien: Elementary school, really. I wrote like a 180-page terrible, terrible novel in fifth grade, where you could tell chapter by chapter what other books I was reading. Because I was just straightforwardly ripping them off. 

So somewhere in somewhere in fourth or fifth grade, around the time that I got access to a computer because a keyboard was just way faster than handwritten, and that was what unlocked the ability to start practicing.

Lynette Bye: And how often did you write between then and the last half dozen years, when you've been publishing more, at least from what I've seen? 

Duncan Sabien: So all through middle and high school, I did a little bit of writing on the side outside of what was required for academic work. Maybe something on the order of like 10 total poems or short stories or essays or whatever, per year. And then ramped up a good bit in college when I started working on a documentary and a novel book. And that's when I think I really made frequent writing a regular part of what I was doing. 

Lynette Bye: What is your writing process like today from that from getting ideas all the way through to publishing? What does that look like?

Duncan Sabien: Yeah, so for shorter stuff, it's often very direct. Like, I'll have the idea. I'll write it that day. I do want to describe my like, second-by-second writing process in a minute, but for shorter ideas, I pretty much just churn them out. Longer ideas or things that I don't have time for that day, I have a list of things to write, I have an ongoing note on my computer of things to write. 

In particular, I noticed that I got a lot more out of that list if, instead of just writing down what I wanted to write, like, “oh, write an essay about x”, I instead started recording the two or three sentence version of how I got excited about it in the first place. Like the “oh, man, it's so cool, because this thing connects to this thing. And I think the two of those together implied this, I should write an essay about it.” I would literally write that down. And that makes it easier to recapture the ideas later. But in general, if there's a long thing, I’ll spend, you know, like one afternoon, sort of plotting and planning and brainstorming, and then after that, just sort of jump in. 

The thing that I think really makes me strongest as a writer, the thing that I think is like most unusual that most other people don't do, is when I'm actually writing the text. Let's say I've written like five paragraphs, and then I sort of peter out, like I run out of steam. What I do for everything I write from like emails to essays to fiction, I'll go back up like two or three paragraphs, hit return a few times, and then I'll just start literally re typing those last paragraphs. Literally just straightforwardly word for word, kind of reading as I read/type, and maybe I'll like start changing one little thing here or there if I notice a sentence is dissatisfying. But usually by the time I get to where I previously petered out, I've caught the flow again, and I just flow forward and write another five paragraphs and then when I run out of steam, I go back up, hit return a few times, start rewriting and flow forward. 

So my very first drafts are usually secretly like second or third drafts in that almost every word has been written twice or three times. But like I do that before I ever get to the end of the piece, and this, this has a lot of benefits like 1) I'm constantly rereading and evaluating the thing that I wrote, but not in like an anxious loopy fashion. It's just a one time. Like, I go back over it one more time and confirm. 2) I'm sort of always practicing finding the thread again. So I get the, “Oh, right, this is extraneous. This is distracting. This is what I really want to talk about” and 3) because I'm reading it aloud in my head as I retype it, I think my writing has a lot more of the verbal quality. It's a lot closer to how it would be said out loud than a lot of people are when they write.

Lynette Bye: Yeah, okay. So when you're retyping it, you're not typing words for word what you wrote before. It's more like you're copying the idea of the paragraph and retyping it from scratch.

Duncan Sabien: I start out typing it word for word. But the second I feel like diverging, I just start diverging. It's easiest to just pick it up by just retyping word for word. But once I catch the flow, then I let it start changing right away. It's also has the side benefit is if I pick something up in the morning, like if I put it away last night, it's very easy for me to warm up and get back into the flow of what I'm writing because I already have a method where I just retype the last three paragraphs and then you're going. 

Lynette Bye: And apart from this, what do you think most distinguishes your writing from, say, the average people on LessWrong? 

Duncan Sabien: My joke answer is that I use a lot of M dashes.

Let's see.

I think this, this might be false. But I think I'm doing something much more like modeling the audience all the time. Modeling different versions of the audience and like, thinking about where it's not going to click and where it's not going to make sense. 

My sense of the average piece on LessWrong is that most of the authors do a really good job of fleshing out the idea itself. Like they get caught up in the concepts, and they explore all the different corners of the concepts, but they lose a little bit of the, like, “What do I do if someone's not intrinsically interested in this concept?”, or “what do I do if they lack a certain prerequisite that I'm taking for granted?” 

And I'm much more, almost anxiously, always thinking because I taught sixth graders and I taught Taekwondo to 40 year old mothers and I just have this sense of like, “Where's this not gonna land? What's gonna go wrong with it?” And that always influences, like, my caveats and my digressions and my explanations of prerequisites and “here, here's the objection.” I'm anticipating all that stuff.

Lynette Bye: Yeah. So you point to these teaching roles as developing this ability to model your audience. But I don't see this as an ability that everyone has who has taught. Do you have an idea of what you did in those roles, or outside of them that really fleshed out this ability to model your audience? 

Duncan Sabien: Yeah. Or at least I have a causal story. I have a chronological story that makes sense when I look back at it. It might be a just so story. But the interesting thing for me when I look back at my own teaching is that I first started teaching within a super rigid framework of a Martial Arts Academy. Where it was like, “okay, Duncan, your job is to teach this exact kata in this exact way. And like everybody should be doing it exactly the same.” And so what was isolated for me was the like, “how do you reach unique people,” if that makes sense? Like there was no variation in the content, the content was just completely set. And it was all about like, “can I convince these skeptical students to do it my way?” “Can I motivate these bored students to keep going?”, “can I explain why you have to hold your fists in this certain fashion up against your ribcage?” Like, I constantly had to be like looking for... 

Sorry, I promise this connects but a quick digression. Kenzie has this thing about CFAR classes, where she's like “there's two different skills.” One skill is the ability to produce a metaphor on the spot that's actually elucidating and clarifying. So Kenzie’s like, “I tried four different ways to make goal factoring work for this person, and it didn't work for them. So I have to come up with a fifth metaphor that will make it click.” And can I make the analogies and connections in that way. And the second skill is having the library of all the metaphors that you previously came up with. So there's generating new ways to make connections and then not forgetting all the old ways to make connections. And this resonated with my teaching, Taekwondo. I was like, “Okay, here's how I would normally explain this to a person. If that doesn't work, here's how I would explain it to like most of the people who are left, and if that doesn't work, then I have to come up with something new.” And practicing that skill, I think lent itself to my writing. Like I think that just got wrapped up in how I think about the audience and how I write.

Lynette Bye: So it sounds like part of what you were doing is modeling the audience in person, gaining that skill before you ever started writing to a remote audience would later read it.

Duncan Sabien: Yeah, I think that's true. And also the fact that there was very specifically a tight feedback loop on modeling the audience itself. Like, that was isolated, and that was the only skill I was building up. 

I didn't have enough of a large sample size to say whether this works or not, but I had a pet theory with my sixth graders that actually, the way to get them to become better writers would be to get them to sort of piggyback on their verbal module, which is like great and mature and like sixth graders are fine at talking and lecturing and debating and so forth. So the thing that I tried doing with them that sort of works, although again, small sample sizes, was I first had them practice transcribing raw dialogue. I would just have them like record themselves having a conversation and then just transcribe it. And then like, “what kind of punctuation do you use to indicate what the speech was really like?” Getting them to get the sense of like, “oh, okay, my writing is just like my talking.” Because a lot of people, I think, when they're starting out writing are really good at talking, but like somehow they just do a horrible job of translating their clarity of thought and speech onto the actual page. So I had them just try like “just get your actual words on the page first, and then go from there.” And this seemed reasonably useful. 

Lynette Bye: College academic writing probably doesn’t help here either. 

Duncan Sabien: Yeah. Yes. Although I don't know. I guess I usually did well with college academic writing. But I think the way that I built up my writing skills was sufficient to get me through a bunch of other kinds of writing as well. But I don't know. I don't know what would happen if you started with that other kind of writing and tried to grow from there. 

Lynette Bye: And going off of that, what do you think were the most useful exercises, mental notes, other things you did to develop specifically your ability to convey ideas through writing?

Duncan Sabien: Yeah. So one is the book Characters and Viewpoint by Orson Scott Card, which is nominally about fiction writing, but actually just is a shockingly good resource for writing in general. I think that literally gave me like a 10% bump in quality when I first encountered it in high school, and that compounds over time. Sorry, that sounds like an answer to your question. Could you repeat the question again? Just so I make sure I’m not wandering. 

Lynette Bye: Yeah, what was the most useful exercises, mental moves, things you did to develop your ability to convey ideas and write well? 

Duncan Sabien: So let me just think for 20 seconds.

Yeah, so there's a small number of things here. One is something like the pyramid model, which we may have talked about in mentorship training. Where like, really actually practice being able to say it in a sentence and in a paragraph and in a five page essay and in like a 20 page thesis. Like practicing the ability to expand all the nodes in the mental workflow. And also to collapse the nodes. If you can only say it in a page, then you need to practice also being able to say it short. But if you can only say it short, then you need to also practice being able to say it in a page. Explicitly practicing the skill of lengthening and shortening, that was big.

Lynette Bye: How did you explicitly practice that?

Duncan Sabien: I would literally just rewrite things at different lengths. So yeah, just take take your paragraph, write a page, take your page, what's the perfect sentence that encapsulates the page? There's also something in there about almost like Gendlin’s focusing or belief reporting, of like, really actually tasting the stuff that you're writing. And looking for the click of “yes, that's exactly it. That's the phrase.” 

Like, I would often when writing character dialogues just rewrite the same thing. They're saying the same thing, but like, they tried to say it and I'm like, “no, that's not right. They're going to regret that version of it in the shower” And then like, have them say a different sentence, and just write the same claim over and over and over again until you really actually get it. Yeah, so practicing looking for the click of like, “I've I have indeed finally expressed the thing” and practicing being dissatisfied with a version that's not quite good enough. Like you don't want to be paralyzed in your writing, you don’t want to be stuck. But knowing as you wrote that sentence that, “okay, that's not good enough. I'm gonna definitely have to come back and try that one again.” 

Lynette Bye: Was there anything that helped you once you identified where you were dissatisfied, that then helps you get to the point of the click “Yes, this is what I actually want to do.”

Duncan Sabien: Just just literally being good at like, explore or a genetic algorithm, just iterating. Sometimes I would rewrite the sentence and just try changing one word and then I would rewrite the sentence and try changing that one word again. And then I'd be like, “nope, this isn't getting me anywhere. Let's try restructuring the whole sentence and starting in a different place.” Just being willing to actually try it 15 times or something. I think there's something where a lot of people will try out... Like, you don't do this for every sentence, right? This is for the crucial sentences. This is the moment in my thesis where I really need to convey exactly the subtle point about social dynamics, or this is a moment in the scene where you can tell that they're not going to make it as a couple or whatever. The crucial moment. 

I think a lot of people are willing to rewrite it three times, and then they just sort of get frustrated and give up. And it's interesting. It's not that much harder to actually try 20 times just for that one sentence. It's sort of like if you go on OkCupid dates. If you go on three okay Cupid dates, you might just be like, “well, I can't find a partner.” But if you're willing to go on like 20 OkCupid dates, suddenly it actually starts working or something. Yeah, there's something about getting people to believe that it will actually pay off if they iterate a little bit longer than they naturally might.

Lynette Bye: Yeah, I am imagining, both for the writing and for that dating analogy that having more things to try than just going out at the same way that you by default do would help. Do you do anything to increase deliberately increase your repertoire of ways to rewrite it? 

Duncan Sabien: Yes, but I think mostly unconsciously. The major thing there is just reading a lot. I will note that as a sixth grade teacher, I think I could have blindly and accurately pointed -- like, if you gave me a student's paper, I could tell you how many books they were reading relative to the other students. And if you showed me how much they were reading, I could tell you what their grade would be on a paper relative to the other students. It's just shocking how strong the correlation is between reading and writing, at least at the sixth grade level. I would imagine that all adults have read a sufficient amount. 

But the more that you read of the type of thing that you want to write, just straightforwardly the easier it is to sort of block from a mental sample of like, “Oh, yeah.” Even if you're just writing like dry instructional text. If you've read 100, dry instructional text, you'll just have much more training data for producing your own dry instructional sentences. Much more data of what works for you, what works for other people, that sort of thing. I think that reading, much more so than trying something crazy experimental with my own writing, just having exposure to more variety of how this might be structured, got me a long way. 

Lynette Bye: I'm curious there. So I think my experience was that reading 100 things might get me to the dissatisfied point. But only by deliberately looking at how the good example structured it, was I able to figure out how to get from dissatisfied to actually phrasing it well. It sounds like you didn’t have to do that step.

Duncan Sabien: Well, actually, no. The thing you just said lands really hard because actually, now that I think back a little more carefully, there are a small number of texts that I did, in fact, very deliberately pour over. Like, “what is the author doing here?” Mostly in fiction. I didn't do this with very many nonfiction, but Hatchet and Enders Game are both books where I would stare at a page and really model and think and be like, “man, how can I write like this? How did he generate these sentences?” I would sort of try to build my own generator out of it and then try it out and then look back at my own work a week later and be like, “does this have the Hatchet nature or not?” And then iterate from there. 

Also, I think I've sort of touched on this a little bit, but reading things out loud. if you're rereading your own writing, it is so easy, if it's something that you wrote just yesterday, to kind of blur your way through it, like you already know what is says. It's hard to really actually read it. Reading it out loud, 100% worth it. Okay, sometimes you read the whole thing out loud, and sometimes you just read the paragraph that you're working on out loud, but actually pausing and putting it into your verbal loop makes a huge difference.

Lynette Bye: And I'm guessing this is reading it out loud, as though you were speaking to an audience and conveying it. Not the whisper under your breath type of thing. 

Duncan Sabien: Yeah, for sure. In fact, if you can put a rubber duck on your desk, that's great. If you have a literal friend that you're like, “Hey, listen, I'm just gonna read this out loud to you and like, you don't even need to do anything. But just like having a real human here will help me a lot.” Yeah, huge difference.

Lynette Bye: Okay, so I was thinking Paul Graham has this post recently about how good writing is as bold as possible, while being accurate. Your writing seems to do a pretty good job of balancing this intersection of bold and calibrated. I'm curious if there's anything in particular you did to develop that skill?

Duncan Sabien: Yeah. Basically, I was really, really bold and then eventually learned to regret it. Like something like that. If you met me 10 years ago, I was much more annoying and expansive and emphatic and like, had basically no caveating at all. And Steph Zolayvar spent five years beating epistemic humility into me, which I'm grateful for. But I think one thing that somebody might try actively is writing a private version. Like, in your first draft, write it to an audience of literally just copies of you. You can just take everything for granted. You don't have to put in the hedge words because everybody takes the hedge words for granted. And you can just be blunt and aggressive and just say the thing straight out. 

Then take the take the version that's like, “Okay, this would work for an exact copy of Lynette” and then be like, “okay, now how do I translate this into like, if I put this up on LessWrong, I won't get yelled at?” And you sort of go back through your own thing you're like, “oh, okay, here I want to add hedge words and uncertainty words, versus over here, what I want to do is actually just be just as blunt as I was, but I will headline it as saying like, Okay, this is me speaking from a part of myself, you know, and like, I'm not trying to be accurate here. I'm just hitting the emotional tone.” And then just proceed with the thing. 

I think the biggest skill here is noticing as you write the sentence, catching the difference between “I think this” or “I believe this” versus “this is true” or “this just is how it is.” You notice yourself writing the latter and then you either change it to the former or later on caveat, like, you know, “that was me being over the top but whatever.”

Lynette Bye: Yeah. You ever have the opposite problem where you were hesitant to be as confident as you could have been?

Duncan Sabien: Not me personally, but that's just like an artifact of my own personal history and my elementary school life and stuff. I certainly think that a lot of people do have that issue. When I'm trying to encourage somebody and I noticed that they're being hesitant or anxious or uncertain or something, I try to get them to connect back to the reason why the thing matters. Like what is the spark of aliveness in this thing? Like, where's your emotion come from? Where's your motivation come from? Take a minute to sink back into why this is so exciting or awesome or scary or whatever. And that usually helps people boot up the expansiveness a little bit. They like, “oh, man, I'm really anxious about like, my own uncertainty and I don't know I can like, convey this accurately enough. I'm like, “Yeah, but you really care about the animals dying, right?”And then they’ll like, think for a minute about like, “oh, man, I really care about the animals dying.” And like that usually loops back in, in a good way. 

Lynette Bye: And as you're conveying these, you tend to do it very extensively through examples. How are you getting these examples? Are you just remembering things that happened to you? Are you putting together stories? Where are they coming from?

Duncan Sabien: I think it's about half and half. This looks back to the Kenzi-claim earlier. One skill is remembering and indexing your examples such that the examples occur to you when you need them to. Actually that Orson Scott Card book I recommended, Characters and Viewpoints, has some exercises on drawing up your examples. And then the other skill is something like being able to just produce examples, and I don't know how I developed the skill. Like I couldn't shortcut you to it. But I would say that just trying it a lot works. We would do exercises. Like, a thing you could do is just like, find a 10-year-old, like an actual 10-year-old, who's willing to humor you for 10 minutes, and just try to explain something really complicated and tricky to them. And then it won't work and it won't work and you'll try another way and it won't work and you'll try another way that won't work. And maybe that's it, maybe you never even successfully explained it to the 10-year-old, but you just got four cycles on trying to come up with examples, and that's useful. 

Lynette Bye: Thank you. Is there anything else that you think someone who's trying to develop their writing skill and ability to communicate ideas clearly, any advice that you would give them?

Duncan Sabien: The one other thing that did occur to me just now is something like, there's the sort of person that goes on a diet and they expect to see immediate results and after like two days of eating right, they actually gained a pound and they just get demoralized and discouraged and just quit right away. I think this is a very natural and human thing to do. 

But there's another way you could approach the dieting, which is like, “Okay, what I'm actually going to do is I'm going to be in this for the long haul. And like, I'm going to try one thing for the first two weeks and just find out if it works. And if it doesn't, I'm going to try something else for the next two weeks and just find out if it works.” And this way, you can sort of like narrativize and contextualize failure into feeling like incremental progress. 

I would really, really advise somebody who's like, “I'm actively setting out to become a better writer”, the first thing I would think is like, “cool, this person is going to get discouraged sooner than they should. They're gonna feel like they're not making progress when they really actually are.” So I want to try to give them some kind of mental frame that will give them the perseverance to keep going. Because after three weeks, they'll just feel like nothing's working. But from my perspective, I feel like that three weeks was already progress. And if they just kept going, it'll be great. So I would try to get people to adopt this scientific mindset of like, “Okay, I have the project of embarking on becoming a better writer. If I tried 10 things, and all 10 of them fail, that is progress. And I will make sure to pat myself on the shoulder, and reward myself for that progress that I made.”

Lynette Bye: I've noticed that when people are trying to take this mindset, they often don't know how long they should keep doing the same thing before they switch to something else. Do you have an idea for trying some of these techniques, how long you would expect them to try it?

Duncan Sabien: Yeah. So it's it varies on whether they've got a tight feedback loop, right? Like if they have an audience that will give them criticism, then I think you can just try it until you learn the lesson. Like you tried this thing and then people complain and then you got a glimmer of insight out of it. Now you can try Something else. If you're more in a vacuum, you don't have that many people giving you feedback, you kind of are on your own, then I would just say, as a number, I would just say try everything five times. 

And the reason I would say try everything five times is because when people try something three times, they pretty much think that they've learned what they're going to learn. And there's usually a little bit left to they can snag after five times. Like, if you have somebody do an exercise. The first time it's hard. The second time they kind of get it. The third time they're like, “Oh, yeah, I get it”. But really, as a teacher looking down at them, you're like, “Yeah, do it two more times to lock it in.”

New Comment
3 comments, sorted by Click to highlight new comments since: Today at 8:00 PM

Strong upvote too. Thank you for sharing this.

There are innumerable resources with techniques like "avoid passive voice", but very few about the writing process itself, which I think becomes a crucial point to focus on if a person wants to advance beyond a "good enough" level of writing ability. Reading about Duncan's process gave me a few ideas on how to improve my own.

I suspect that this why writers' workshops help people improve--they allow writers to debug each others' writing processes. 

Strong upvoted: This is great! This post's form as a transcript made it engaging to read, and I feel like I learned from reading it.

In particular, both "rewriting to catch the flow" and "practicing the ability to expand/collapse all the nodes in the mental workflow" are interesting ideas I'd like to try in the future.

In general, the whole post synthesizes some different ideas I've heard about writing in an informative way: (sort of a topic dump) Writing as talking, thinking about the audience, writing down ideas for future expansion, rewriting effectively, noticing (both when a sentence isn't "perfect" yet, and whether you just believe a sentence vs. knowing it's true), and practicing writing as a long term habit (and therefore short term improvement might not be significant).

Thank you for sharing!

the thing that I think is like most unusual that most other people don't do, is when I'm actually writing the text. Let's say I've written like five paragraphs, and then I sort of peter out, like I run out of steam. What I do for everything I write from like emails to essays to fiction, I'll go back up like two or three paragraphs, hit return a few times, and then I'll just start literally re typing those last paragraphs. Literally just straightforwardly word for word, kind of reading as I read/type, and maybe I'll like start changing one little thing here or there if I notice a sentence is dissatisfying. But usually by the time I get to where I previously petered out, I've caught the flow again, and I just flow forward and write another five paragraphs

This sounds really interesting. I will write another comment once I've tried this a few times.