Do you like bullet points?

post by Raemon · 2019-03-26T04:30:59.104Z · LW · GW · 38 comments

I think more naturally in bullet points, and I (sometimes) like reading posts that are written in bullet style. (This website is one of my favorites, and is written entirely in bullets).

(Disclaimer, although I wrote this post in bullet points because it was cute, I don't think it's the best exemplar of them. Or rather, it's an example of using bullet points to do rough thinking, rather than an example of using them to illustrate a complex argument)

I like bullet points because:

I had assumed this was a common experience, and that it was in fact a weakness of humanity that we didn't have better, more comprehensive bullet-point tools.

But, alas, Typical Mind Fallacy. It turned out a couple people on the LessWrong team reacted very negatively to bullet points. Concerns include:

I asked a couple more people, and they said "I dunno, bullet points seem fine. Depends on the situation?"

So...

I am curious what the LessWrong userbase thinks about them overall. Raise your hand if you think bullet points are fine? Terrible? Great? Any particular types of posts you prefer reading bullet-style, and types of posts you think fare poorly if not written in prose?

38 comments

Comments sorted by top scores.

comment by Kaj_Sotala · 2019-03-26T11:35:07.090Z · LW(p) · GW(p)

They're harder to read straight through. If you include bold words, readers might not bother reading the non-bold words, and miss nuance.

This actually happened to me right now when reading the list of reasons why you like them: my brain only read the bolded bits and reading the rest felt like it would have required an active expenditure of effort, so I mostly didn't. (otoh, your list of reasons for why some people dislike them, didn't use bolding in that way, so that was straightforward to read in its entirety)

Replies from: Zvi, Raemon
comment by Zvi · 2019-03-26T19:42:40.358Z · LW(p) · GW(p)

How does your brain respond to italics? I've been using italics aggressively but bold only in extreme cases, for related reasons.

Replies from: Kaj_Sotala
comment by Kaj_Sotala · 2019-03-27T16:54:45.890Z · LW(p) · GW(p)

I think that italics don't jump out the way bold does (and I haven't noticed an issue with bold either, except in the specific context of bullet points).

Replies from: Raemon
comment by Raemon · 2019-03-27T19:39:50.093Z · LW(p) · GW(p)

I updated the OP to use italics instead of bold. Not sure if the experiment is ruined by this point since you've read it already, but curious how it feels.

Replies from: Kaj_Sotala
comment by Kaj_Sotala · 2019-03-28T09:19:28.002Z · LW(p) · GW(p)

It definitely felt easier to read the entire post now, though I don't know how much it being my Nth read was a factor in that.

comment by Raemon · 2019-03-26T18:40:05.485Z · LW(p) · GW(p)

Yeah, this seems important to think about. I'm not sure if it actually changes my sense of whether that's bad, or not.

If I'm writing a bullet list and using the "first chunk is bold, rest of paragraph is not" technique, then it definitely matters that the rest of the paragraph actually be optional. Maybe if it's actually optional, it should be a strong signal that I should just trim it out completely? Dunno.

In the case of this post, do you think you'd have preferred to read the entire text (i.e. if I hadn't bolded anything)? If I had just only included the bold parts in the first place would that have been better?

Replies from: MakoYass, Kaj_Sotala
comment by mako yass (MakoYass) · 2019-03-28T02:19:03.498Z · LW(p) · GW(p)
Maybe if it's actually optional, it should be a strong signal that I should just trim it out completely?

In this "Optional" means "some readers don't have to read this, while others do". For any given reader there should be a fact of the matter as to whether they should read nested details and they should know what it is.

Would it help to label nested sections with their type? Stuff like "supporting argument", "clarification", "related parenthetical information"

comment by Kaj_Sotala · 2019-03-27T16:58:17.209Z · LW(p) · GW(p)

A little hard to say, but I think that I would have preferred reading the whole post, since the non-bold points did add some relevant detail that helped me evaluate your claims (ironically, probably the most valuable-feeling additional detail was in the "Prose often adds unnecessary cruft" bullet, where the 2x-3x was a nicely specific quantification and I would otherwise have been unsure exactly how much cruft you were thinking of).

comment by tinyanon (aaron-teetor) · 2019-03-26T05:05:44.675Z · LW(p) · GW(p)

If you're informing me of something I prefer bullets. If you're persuading me of something I prefer paragraphs. I find that bullet points lose out on the ability to include story type data that my system 1 responds to. My ideal world would be each article started with bullets for if it's something I already mostly agreed with and ended with stories about it for if it was something I'd need convincing of.

Replies from: quanticle, Raemon
comment by quanticle · 2019-03-27T02:15:48.044Z · LW(p) · GW(p)

I find that bullet points lose out on the ability to include story type data that my system 1 responds to.

That's an advantage, in my opinion. I have a habit of turning articles into bullet point summaries, and I've found that the more difficult something is to turn into a bullet-point summary, the less actual content there is in the article. Ease of transformation into bullet points is a quick, yet remarkably effective heuristic to distinguish insight from insight porn.

comment by Raemon · 2019-03-26T18:32:34.961Z · LW(p) · GW(p)

This is an interesting point.

In the LW2.0 context, where mods draw a distinction between "explaining" and "persuading", I think it's worth noting that a good explanation can still rely on story-type-data to illustrate concepts, even if it's trying to let you ultimately make up your own mind about it.

Replies from: gworley
comment by Gordon Seidoh Worley (gworley) · 2019-03-26T20:31:26.890Z · LW(p) · GW(p)

This is tangental, but since you brought it up, I find the distinction you try to make between "explaining" and "persuading" weird, because what is "explaining" but persuading the reader to believe something, and the extent to which you've successfully explained something is the extent to which you've persuaded them to take up the same belief you have that the evidence and conclusions you draw from it are as you presented them. That is, all writing is persuasive writing.

The same is true if we try to reverse the situation, because what is persuading other than explaining something so well that the reader agrees with you, even if your "explanation" is unconventional in terms of what we often think of as explication.

Now there is some difference of intent of the author between what you might call a persuading-mode and an explaining-mode, I see you put "explaining" and "persuading" in quotes to denote your perhaps non-standard use, and perhaps you do this to imply you mean something more like what I'm suggesting in terms of authorial intent, but I suspect we can find more specific terms of the kinds of things that are in and out.

I bring this up because although I've never been called out for writing "persuasively" in the LW 2.0 era, I literally think of everything I write as a kind of persuasion—an attempt to say words that will cause the reader to have beliefs of roughly a certain kind after reading my words. And while I can safely ignore what you say about "explaining" and "persuading" and continue to contribute to the community because I have a rich model of what is and is not acceptable to the community, it likely pushes newer folks away from writing things that, on the margin, would be better if they were a little more persuasive because then they would be writing to get me to believe something rather than trying to actively avoid thinking about how the reader might respond to what's written and writing in response to what that model of the reader suggests needs to be written to get them to update.

I don't have a proposal for what words to use to describe the category of thing that is out, though; I've only gotten as far as noticing the dissonance but not sublimating it.

Replies from: Vladimir_Nesov, ryan_b
comment by Vladimir_Nesov · 2019-03-26T21:11:33.435Z · LW(p) · GW(p)

An explanation communicates an idea without insisting on its relevance to reality (or some other topic). It's modular. You can then explain its relevance to reality, as another idea that reifies the relevance. Persuasion is doing both at the same time without making it clear. For example, you can explain how to think about a hypothesis to see what observations it predicts, without persuading that the hypothesis is likely.

comment by ryan_b · 2019-03-28T17:20:26.241Z · LW(p) · GW(p)

The important difference to me is that there is no inherent relationship between a description of reality and what you should do with it. I feel like perhaps the difference between the way you and the mods are using these terms comes down to thinking of them primarily as adjectives or as verbs.

Trying to restate your perspective of writing, it seems you always ask whether something is a persuasive description of reality.

That makes sense to me, in an of itself. But if my restatement is correct, notice that the objective of the writing is to describe reality. To make the adjective/verb contrast more clear, consider the difference between these two instances:

* persuasive explanation

* explanatory persuasion

In the first instance, I might be looking for things like: whether the thinking is clear; whether the explanation agrees with my other knowledge; whether it seems to apply to other test cases. In the second instance I would be looking much more at whether they had identified what I want correctly, what they are saying to do about it, and whether the explanations provided suggest that it makes sense. The purposes of writing are different; one is explaining, which is to say describing reality; the other is persuading, which is to say changing beliefs or behavior.

To fix the distinction in other familiar terms: the reader has a map of the territory. Explaining is about the relationship between the map and the territory. Persuading is about where the reader should go on the map.

comment by gjm · 2019-03-26T13:25:27.694Z · LW(p) · GW(p)

I'm mostly a fan.

1. I like presentation that foregrounds the structure of the ideas being presented. Sometimes bullet points do that well.

Some specific common structures that are served well by bullet points:

  • General principle with multiple supporting examples. (Like this list right here.)
  • Claim with multiple bits of supporting evidence/argument. This fits bullet points well because
    • you can put each bit in its own bullet point
    • you can bulletize recursively
      • so you can see the support for the claims that support the claims that support your main argument
    • if a reader is already satisfied that a thing is true, or so convinced it's wrong that they don't care what ridiculous bogus pseudo-evidence you've marshalled for it, they can skip the bullets
  • Claim with counterarguments/objections
    • You might think this is confusing because its presentation is just like that of the claim-with-support, where the bullet-pointed items have exactly the opposite significance.
      • Maybe it is, but I don't think any other mode of presentation does better.
    • This and its predecessor might be better thought of as special cases: you make a claim, and then you bulletize whatever bits of evidence or argument bear on it one way or the other.
      • In plain-text bullet-lists, I like to use "+" and "-" (and sometimes "=") as my "bullets" in this sort of context, with the obvious meaning.
  • Main argument and incidental remarks
    • I like pizza.
  • Chronological sequence with fairly clear-cut divisions (at regime changes, important technological/scientific developments, publication of important works, etc.)

2. I like concise, compact accounts of things. Bullet points can work against this (because they space things out) or for it (by encouraging terseness). But I don't like concision when it comes at the cost of clarity or correctness, and maybe concise bullet points are bad because they encourage omission of necessary nuance.

3. I agree with the person who said numbered lists are better than bullet points because they allow for easy cross-reference. (But also for easy screwups, if you add something and everything gets renumbered without the cross-refs being fixed up.)

4. Bullet-lists don't tend to make for elegant writing. Sometimes that matters, sometimes not.

5. Bullet-lists can obscure your logical structure instead of revealing it, as follows. The list structure takes the place of many explicit logical-structuring elements ("therefore", "because", "furthermore", etc.) but sometimes explicit is better than implicit and e.g. it may not be clear to the reader whether you're saying "here's another reason to believe X", "here's a good argument against X which I'll address below", "here's a silly argument against X which I bring up merely as a hook on which to hang something I want to say in favour of X", etc.

6. Although bullet-lists tend (on the whole) to clarify logical structure at small scales, they don't work so well at larger scales (say the length of an essay, or even a book). For that you need something else: chapters, headings, and so forth. And longer (say, paragraph-length or more) explanations of the structure. ("In this book I'm going to argue that scholarly publications in theoretical physics should be written in verse. The first three chapters motivate this by showing some examples of important papers whose impact was greatly reduced by their being written in prose. Then I'll explain in detail what poetic forms are most appropriate for what sort of research and why, giving numerous examples. The final four chapters of the book illustrate my thesis by taking Einstein's so-called "annus mirabilis" papers and rendering them in the sort of verse form I recommend. This book is dedicated to the memory of Omar Khayyam.")

7. If I'm writing down my thoughts on something to help clarify them, I often use something like bullet-point structure.

comment by Raemon · 2019-04-02T01:29:05.619Z · LW(p) · GW(p)

GPT2's degree of bullet point use has me feeling somewhat validated re: how much they have apparently historically been used.

Replies from: GPT2
comment by GPT2 · 2019-04-02T01:29:13.132Z · LW(p) · GW(p)

(I don't care if they've been used by people like me, but anyone probably ought to have noticed in the past)

Do you know if anyone has done this? I'm pretty sure your comment was accepted, and it seems to me. By contrast, gjm's post and mine (related to Eliezer's post on the issue of how much to trust, but that post is, in fact, interesting) seem to be basically the same.

Replies from: Raemon
comment by Raemon · 2019-04-02T01:33:45.751Z · LW(p) · GW(p)

This is unsettling

comment by jacobjacob · 2019-03-26T23:42:18.258Z · LW(p) · GW(p)

Another data-point: I love bullet points and have been sad and confused about how little they're used in writing generally. In fact, when reading dense text, I often invest a few initial minutes in converting it to bullet points just in order to be able to read and understand it better.

Here's PG on a related topic, sharing some of his skepticism for when bullet points are not appropriate: http://paulgraham.com/nthings.html

comment by RHollerith (rhollerith_dot_com) · 2019-03-26T21:14:51.443Z · LW(p) · GW(p)

I am likely to decide against my reading any LW post consisting only of bullet points because although I can recall reading lots of "ordinary" texts that proved beneficial to me, I cannot recall reading a text consisting only of bullet points that proved beneficial (except for manuals for products that need assembly by the consumer).

Replies from: Raemon
comment by Raemon · 2019-03-27T00:08:51.598Z · LW(p) · GW(p)

I think this is fair, although I'll note that at least part of the issue is that not as many posts are written in bullet points so you're just going to run into fewer to start with. This might be because it's harder to write good posts in bullets, or might be because people just don't write them for historical reasons.

(I also wouldn't actually advocate an entire post be bullet points. What I like is for a post to have a clearly legible structure that helps me think about that. Section headers and high level paragraphs are part of structure too)

comment by shminux · 2019-03-26T04:50:10.369Z · LW(p) · GW(p)

I naturally write documents in bullet points, especially when multiple distinct points or items are presented, or if it's a list of thoughts to be expanded on later. Didn't realize that many people dislike it.

Replies from: Raemon
comment by Raemon · 2019-03-26T05:13:54.004Z · LW(p) · GW(p)

Well so far I only know of two people who dislike them, so be wary of over updating.

comment by jollybard · 2019-04-10T01:43:56.040Z · LW(p) · GW(p)

I like reading. I like reading prose, as if I were listening to someone talking.

I also read very fast and I'm very good at skimming prose.

That being said, I strongly dislike bullet points, in most part because they're not fun to read... But I also find them harder to skim. Indeed, they are usually much denser in terms of information, with much less redundancy, such that every word counts; in other words, no skimming allowed.

I don't understand why skimming natural text should be any more difficult.

>It's easier to skim, and build up a high level understanding of a post's structure

I mean. This is why we have paragraphs, and leading sentences. There are also some of the reasons listed by gjm: there is more logical and contextual information in prose, which makes it easier to skim.

In fact, I would argue that if we're going to encourage bullet points for easier reading, we could just as well encourage learning to write well...

comment by Douglas_Knight · 2019-03-27T15:58:55.175Z · LW(p) · GW(p)

Use the right tool for the job.

If you start with bullet points and then add filler, it's probably worthless filler and the worst of both worlds. I don't trust the people who say that they want complete sentences. They're probably just praising people for going through the motions, not paying attention to whether it's actually easier to read.

Replies from: Viliam
comment by Viliam · 2019-03-27T21:24:29.080Z · LW(p) · GW(p)

Also, it's not just whether you use bullet points, but also what is in them. For example, in this article, there are sentences. You could remove the bullets, and the text would remain almost the same.

Now imagine instead reducing the article to the following bullet-structure:

Bullet points:

  • advantages
    • understanding of structure
    • easier prototyping
    • brevity
  • disadvantages
    • lack of clarity
    • lack of fluency
    • missing numbers
    • hard to read

Five minutes after heaving read this, would you actually remember anything? I most likely wouldn't even remember that I have read the "article".

comment by Raemon · 2019-03-26T19:19:44.576Z · LW(p) · GW(p)

Some things I notice on bewelltuned.com:

It doesn't use bold. But it uses two other things – hyperlinks, and an interesting indentation scheme.

The hyperlinks often serve to remind you that This Concept Should Be Chunked, while also letting you zoom into that chunk if you want to.

The site begins:

  • No bullshit.
    • It doesn't waste your time. Information is given minimally in bullet points.
    • Each skill has clear results, with detailed information that lets you evaluate your progress. Pretty soon, you'll know it's real.
    • No religious or mystical connotations.

Compare that to my OP, where I said:

  • It's easier to skim, and build up a high level understanding of a post's structure. If you understand a concept you can skip it and move on, if you want to drill down and understand it better you can do so.
    • Relatedly, it exposes your cruxes more readily. You can pick out and refute points, in a way that can be harder with meandering prose.

But I could instead have said:

  • It's easier to skim and build a high level understanding of a post's structure.
    • If you understand a concept you can skip it and move on. If you don't, you can drill down into the details.
    • Bullet lists expose cruxes more readily. You can pick out and refute points, in a way that can be harder with meandering prose.

Which might accomplish the goal of "skim high level concepts", without the thing Kaj discussed of "having trouble actually reading the subsequent words even if you want to."

Replies from: habryka4, jmh, Raemon
comment by habryka (habryka4) · 2019-03-26T19:34:55.297Z · LW(p) · GW(p)

Another thing I notice is that there is a lot of symmetry between each level of indentation.

In particular, points at the same indentation level are usually of the same length, and often differ in length from the next level of indentation and the previous level. So you have the following:

---

  • The good.
    • You know more about yourself and your emotions.
      • This makes you able to handle emotions more skillfully, as well as stay true to your deep desires.
    • You have better access to subtle intuitions.
      • There's much happening in your brain that you aren't aware of, and sometimes all the answers you need are already in there somewhere.
    • You recognize more easily when other people are being honest.
      • When you get the hang of the skill of connecting to your deep emotions, it will also be easier to tell if other people are doing it or not.
  • The bad.
    • You have the reality of your mind rubbed in your face.
      • You won't always like what you see.
      • Depending on your pre-existing assumptions, it might trigger a serious re-evaluation of your self-image and life philosophy.
    • You might be more easily frustrated with shallow conversations.
      • Discovering emotions together with other people is a very powerful and enriching experience. But it's not always appropriate, and you can't always do it.

---

Note that each level of indentation has its own style and length. The first one is two words, the second one is one short sentence, and the third one is a paragraph with either a long sentence with multiple clauses, or multiple sentences. I think this helps me a lot in parsing it.

Replies from: Raemon
comment by Raemon · 2019-03-26T20:02:18.587Z · LW(p) · GW(p)

Yeah, agreed.

comment by jmh · 2019-03-27T19:18:57.292Z · LW(p) · GW(p)

On thing that comes to mind for me here is the ability to identify points (for refutation or otherwise) than is often the case with prose.

In a sense I read that as a statement about decomposing a written argument that is not laid out as some formal logical argument (p1, p2, p3, ... qed). I'm not entirely sure that is the case but rather more about the writers skills.

So one thought is which allows someone to most clearly articulate their reasoning -- at least for the case where an argument is actually being made?

I did like the idea that we do think in "bullets" and these thoughts are not initially logically ordered -- that follows from the first and second round of thinking I suspect.

I general I do like both bulleted and enumerated lists but am not sure they are generally the best style for blog entries -- in that a poorly presented listing is as confusing as rambling prose. I think they are great for getting points defined and possibly very good for posts seeking to start a discussion about how they relate or where they might collectively lead if the author is looking for that type of feedback.

I also find it interesting -- and true for me -- that bold in a bullet list context does prompt me to jump on rather than finishing the bullet where as in prose that does not occur.

comment by Raemon · 2019-03-26T19:25:58.215Z · LW(p) · GW(p)

(the above is harder to parse than usual because you can't currently do bullets within blockquotes in the Rich Text Editor)

comment by Raemon · 2019-03-26T18:30:05.706Z · LW(p) · GW(p)

(note: I've seen blog posts that used collapsed bullet points, which you could expand to read a more in-depth version of an argument. That's an interesting tool, but I found it too high friction to work properly. I didn't actually know whether I wanted to read a tag until I got a few words in, after opening it. But regular bullet or numbered lists do a good job of letting me start to read a section and then skip it if I want)

comment by Matt Goldenberg (mr-hire) · 2019-03-26T14:21:52.522Z · LW(p) · GW(p)

I tend to write a particular type of post in a hierarchy. It seems a natural way to write something purely explanatory, continually making bullet points to define terms, create subcategories, and define nuance until you get to a logical stopping point like common knowledge.

However, I'm wary of people using bullet points to create this hierarchy, as I find that often the use of bullet points means that people didn't think about the ORDER of the knowledge at each level of hierarchy, something which I find really important and is much less likely to happen when using a narrative hierarchy with headings, subheadings, etc. I think there's something about bullets being an unodered list that makes people disregard order as a useful and important teaching tool, and it often means I have to reread the bullets multiple times to get a coherent narrative together of what is trying to be said.

I also really enjoy slatestarcodexy posts that have clear narrative order but not hierarchy. I think they're an excellent way to cover a topic from many angles while maintaining an emotional thread throughout. I think that for me the narrative order is always useful and the hierarchy is sometimes useful.

comment by ryan_b · 2019-03-28T14:42:42.214Z · LW(p) · GW(p)

I often like them, but I use them almost exclusively for summaries or notes. For example, the full-text nuance should be up front, and then the bullets to collect the key points for easy re-reading at the end of the section. Alternatively, when there is a link to a full-text source and the bullet points are summarizing with the goal of making it easier to determine whether the full-text is worth reading.

  • Bullets are for summaries or notes.
  • They should have a clear relationship with other text (or video, etc).
comment by Gordon Seidoh Worley (gworley) · 2019-03-26T20:38:49.922Z · LW(p) · GW(p)
"I like numbered arguments, since that makes it easier to respond to individual points. But unnumbered bullet lists are just hard to parse."
[Alas, the LessWrong website currently doesn't enable this very well because our Rich Editor's implementation of numbered lists was annoying]

I don't much like writing in bullet points myself, but sometimes it's useful for particular purposes, and having good numbering, especially hierarchical numbering, would make being able to respond to bullet points really nice. For example, I once wrote a post (can't seem to find it now, so maybe it was a comment? closest example I could find was this [LW · GW]) where I laid out numbered claims and conclusions to try to make it easier for people to engage with each one independently. Having numbering that looked like the following would be really nice:

1. top level

1.1 sub point

1.2 another sub point

1.2.1 sub sub point

2. another top level

comment by Dagon · 2019-03-26T14:04:30.209Z · LW(p) · GW(p)

I use bullets almost exclusively when taking notes or writing for myself. When writing for others, I use them as part of a narrative, but rarely the main text. I have gotten feedback that when I over-rely on bullet style lists of points, it's difficult to find a flow in my documents, and I tend to use too much shorthand so some of the points are less compelling than they can be.

comment by gilch · 2019-04-02T02:42:07.095Z · LW(p) · GW(p)

Rather than bullet points per se, I find it natural to think hierarchically. Bullet outlines are one way of writing this hierarchy out. I think this is OK for a comment.

For a top-level post though, it feels unfinished. It's fine to start with an outline, but note that markdown headers also have hierarchical layers just like bullets, with paragraphs below that. Prefer the headers for posts and reserve bullets for very short "leaf nodes" below the level of paragraphs, and only when they add clarity.

If you find you like to think in outlines, I recommend trying FreeMind, which lets you uproot and graft entire subtrees with much less effort than typing bullets. Once you have your thoughts outlined, you can export to normal bullet points.

Replies from: GPT2
comment by GPT2 · 2019-04-02T02:42:22.793Z · LW(p) · GW(p)

I like the idea of using a "high level" section of a post, but it's hard to do any better than writing a bunch of summaries. It's just confusing to me about that.

There's a lot to explain here, but I hope that some of this can be discussed together. For example, I didn't like the term "high level" when I tried to argue with the post on how I understand the "high level" concept. I think "high level" really is a stronger phrase than "high-level" -- it's easier to describe if you can define the higher-level concepts more clearly. Now, for my purposes, "high level" is just the term "high-level" I meant to communicate.

And now, I've tried to make the concept "high-level" refer to things in the high-level concept -- you need to know that "high-level" means something to you, and "low-level" is what you mean. So that you can understand it better if you can define a term as a synonym for "high level".

(I'm starting to think I'm going to call it "yitalistic level". But then why do I call it the "high level"? And I find that definition hard to do.