In favour of terseness
I like posts that are concise and to the point. Posts like that maximize my information/effort ratio. I would really like to see experienced rationalists simply post a list of things they believe on any given subject with a short explanation for why they believe each of those things. Then I could go ahead and adjust my beliefs based on those lists as necessary.
Sadly I don’t see any posts like this. Presumably this is because of the social convention where you’re expected to back up any public belief with arguments, so that other people can attempt to poke holes in them. I find this strange because the arguments people present rarely have anything to do with why they believe those things, which makes the whole exercise a giant distraction from the main point that the author is trying to bring across. In order to prevent this kind of derailment, posters tend to cover their arguments with endless qualifications so that their sentences read like this: “I personally believe that, in cases X Y Z and under circumstances B and C, ceteris paribus and barring obvious exceptions, it seems safe to say that murder is wrong, though of course I could be mistaken.” The problems with such excessive argumentation and qualification are threefold:
The post becomes less readable: The information/effort ratio is lowered.
It becomes much more difficult to tell what the author genuinely believes: Are they really unsure or just trying to appear humble? Is that their true objection, or just an argument?
Despite everything, someone is STILL going to miss the point and reply that sometimes killing people is ok in certain situations, and then the next 100 comments will be about that.
By contrast, terseness makes posts more readable and makes it less likely that the main point is misunderstood. So if we as a community could just relax the demand for argumentation and qualification somewhat, and we all focussed on debating the main points of posts instead of getting sidetracked, then perhaps experienced rationalists here could write nice and concise posts that give clear and direct answers to complicated questions. Instead, some of the sequences are so long and involve so many arguments, counter-arguments and disclaimers that I feel the point is lost entirely.
Short posts are bad at crossing inferential distances. They limit your readers to those who already know, or almost know, what you’re talking about. (Also see: Explainers Shoot High. Aim Low!)
For example, if you were writing a scientific paper and going for extreme terseness you would give the data then stop. There’d be no abstract, no explanation of previous work and surrounding context, and no guiding towards conclusions. It would be an incredibly inefficient way to convey knowledge.
Another example is computer programs. Programmers try to avoid repeating themselves, but not to the point where they attempt to hit the Kolmogorov complexity (code so unreadable it’s unprovably correct!) or omit all tests (which are by definition redundant).
A more personal example is a post I made recently: Building your own Quantum Fourier Transform. I did not explain how to use the little quantum circuit inspector in enough detail, and so all the feedback was “I don’t get it”. I try to always keep in mind that saying things like “a superposition assigns a weighting to all the classical states” just doesn’t translate, you have to break it down every time, but I didn’t pay attention to that nagging doubt.
Unfortunately, long posts are also bad at crossing inferential distances. One reason is that if a post is long people will have forgotten the first half of the post by the time they reach the end. And the longer a post gets the more likely it is people will focus on a small detail instead of the bigger picture. Not to mention that it’s more difficult and time-consuming to write long posts, so quality will suffer.
Bite-sized posts that have only one thing people should take away from it are much less likely to be misunderstood. And they can still can still teach people something or change the reader’s mind, if only by a little bit. And if there is a missing dependency (e.g. how does a quantum circuit inspector work?) then that can be the subject of another bite-sized post.
You can of course persuade people by throwing lots of text at them. For instance long form sales copy converts insanely well, but there’s some dark-side epistemology going on there. Ultimately people have to stop, think, and process all the inferential steps between A and B. There’s no way around that, so you can’t bridge a big gap in a single long post.
As for the rest I think it’s mostly a matter of good writing. Use simple words instead of complex ones. Avoid adjectives and passive voice. Prefer short sentences. Make it clear what you’re arguing and why. And keep it terse and to the point.
The synthesis obviously is writing sequences of terse posts. If you need to explain a quantum circuit inspector, do so in its own post.
Clearly there’s a balance to be struck here: You want your posts to include everything that’s necessary for your audience to understand it, but no more. “Perfection is achieved not when there is nothing left to add, but when there is nothing left to take away”, as they say. And clearly, the more technical a post is, the greater the risk that it won’t be understood.
However, when it comes to your example of scientific papers, I definitely do feel that some academicians should be more concise! For economics papers, often about 80% of the paper is simply repetition of what other people wrote, sometimes solely to show that the author did their research. Maybe 5% is original input, and the rest is repetition (again, less so for technical papers). Philosophers tend to be even worse!
Scholarship is a virtue, and placing what you did in the wider context is a very important part of pushing science forward (civilization could advance faster if people were more aware of other stuff out there—how many times have people reinvented graphical models?)
Signal to noise depends on the quality of the author—there are plenty of dense papers in academia, in all disciplines.
Personally, I find the Sequences, both individually and collectively, to be a model of clear writing. But they’re not at all terse, are they? Yvain writes long blog posts that I’d put above Eliezer’s for quality of writing. Moldbug writes at insane length, and yet always very readably. But then, some other long posts I find myself thinking, oh, get on with it, get to the point already, whatever the point is. Or they just say the same thing in ten different ways, as if adopting Paul Graham’s advice on how to seem articulate. I can’t give examples because I don’t pay much attention to those.
Short vs. long is not the issue. Structure is the issue.
I’m not interested in short and clear answers to complicated questions. Or to put that more tersely, I’m not interested in answers on their own. My reaction will be, “That’s nice, but how did you get there? Why should I care about your bald pronouncement?” It’s like being given a purported map of buried treasure that is just a blank sheet of paper with an X marked on it.
In that the individual sentences are well-formed. But these days, that’s not enough.
(I note that the Urbit manual shows him actually writing with conciseness and precision—so he can in fact do better, he just doesn’t want to—though the Urbit security document goes a bit Moldbug two-thirds of the way down.)
It could be a matter of taste. Neoreactionary writing may best be considered a species of poetry, art for its own sake.
A few (concise) notes:
Being concise or verbose is a false dilemma. You can structure your post so that it can be both. (I posted about it here years ago.)
You assume that if “terseness makes posts more readable” to you, then it applies equally to everyone. It doesn’t.
Crossing an inferential gap is harder in a short post, unless you are an amazing writer.
If you scan through the site carefully, there are plenty of quality short posts here.
An extra note:
In the quote, the qualification is unnecessary. Ceteris paribus, it’s usually harder in a short post, regardless of general writing skill.
I’d be interested in seeing a trial run of this kind of post. Do you have an interesting belief that you think you could communicate in a compact manner?
Here is a try: If I had to summarize my view on free will it would be: Free will is what deterministic decision making feels like from the inside. Humans are capable of considering alternative courses of action, eventually deciding what to do. The process of choosing occurs deterministically and is what we call free will.
If this were my first introduction to this idea, I would not have a very good idea of what this means. Unpacking “what deterministic decision making feels like from the inside” means would be absolutely necessary for this, and would probably require an example to explain.
Yes, free will is a good example of a subject that is often made ridiculously complicated. (Edit: Removed the bit that was wrong). Here’s my quick (edited) attempt on free will:
I was actually thinking about writing and posting a Mini Meta-ethics Q&A in this style. There is so much overly convoluted writing and general confusion on the subject that I feel there would be added value in this. What do you (and others) think?
If that’s representative, I think what you’ll end up writing is something that people who already share your reasoning and/or conclusion can recognize and validate/feel validated by, and everyone else will gain nothing of value from.
Is that intended to be a summary of what Eliezer wrote, or a summary of your own view? I don’t recognise it as a summary of Eliezer’s writing on the subject, which is much more than a single article. (Non-spoiler posts; spoiler posts).)Either way, there is nothing in that summary that would move me to update in any direction.
Oops. That’s what I get for basing a reply on memory alone. I have to say, 11 blog posts does seem kind of ridiculous for what Yudkowsky calls “the easiest ‘hard’ question”. Not sure why I thought it was only one.
He isn’t a terse writer. And I’m not sure if you want terse expression or simplification of concepts themselves.
Free Willis not conscious volition. There is a puzzle about the co existence of free will an determinism. There is no such puzzle about conscious volition. A few minutes uses ago, I made choice between chicken curry and beef curry. I chose chicken. Could I have chosen beef? You simple answer doesn’t tell me.
Sure you could have, in the sense that IF you had desired beef more strongly, you WOULD have made that decision. The fact that it’s deterministic does not make it any less your choice. I honestly don’t think answering this requires more than one line, since it’s in principle a really simple issue. Convincing someone of this is a lot harder and would take much longer of course, but that’s not really the aim of a terse post. The fact that you don’t feel convinced doesn’t mean the question hasn’t been answered.
Could I have desired beef more strongly?
It’s not a simple issue. It seems intuitive to me that I have ownership of actions that I originate in a way that I don’t over events that were originated by the Big Bang
Answering questions catechistically is easy. Justifying an answer as being the one true answer, and meeting objections is difficult.
The world does not abound with simple answers to complex problems, because they are hard to achieve—genuinely.
The typical failure modes are:
1 answering a different, easier question.
2 taking sides on an issue without sufficient justification, ie coming up with doctrine.
A lot of the problemisthe ambiguity of “answer” between “solution” and “response to a question”.
I find this bit incredibly confusing:
I pattern-match this to attributing agency to evolution?
Also, there is an obvious distinction between your deciding an action freely and affecting a decision (second and third sentences).
I appreciate the example, but I think the terseness here significantly lowers the informational value.
Admittedly, a one sentence explanation of identity was always going to be confusing (even assuming I didn’t screw it up)
Ah, no. I was saying that if a conscious brain didn’t do anything useful, we wouldn’t have a conscious brain. Survival of the fittest and all that. Therefore, the claim that our conscious selves have no control over our actions at all is silly. (unless our consciousness is merely an accidental by-product of something that does have a function… but that seems unlikely) There is indeed a difference between deciding an action freely and affecting it, which is why I ended with “it can be said that we have limited free will”.
I appreciate the input you and others are giving me here. I agree that covering Free will in one paragraph is too optimistic. One post instead of 11, however, still seems quite reasonable to me.
Determinism means that you were not the creator of the circumstances that led to the decision you made. So, in a sense, it does make it not really “your” decision.
It’s really hard to convincingly counter all possible objections in just a few lines. The above was meant to give an instance of a hypothetical list of “things rationalists believe and why”, not something that could replace an actual sequence . Obviously you one cannot hope to clear up all possible confusion in so short a post.
If I were to make a serious terse post about free will, I’m thinking I would at least have to take some time to try and dissolve the concepts of ‘identity’ and what it means to be ‘responsible’ first. This would still keep the information/effort ratio high by explaining multiple related concepts simultaneously. I’ve edited this in above to make it a little clearer, but there’s only so much space to work with...
One reason people write long posts is that they want to anticipate objections (that may or may not be nit-picky). The reason for this is that having to modify your views as a result of the discussion is generally seen as a failure.
The solution to this is to see posting and discussion less as a competitive game the purpose of which is “winning” and more of a cooperative game the purpose of which is learning. That goes for both posters and commentators, obviously.
That said, you often need to explain difficult ideas in detail in order for others to understand, as noted below.
The other reason to contain as much of the arguments and counter-arguments as you can in the original is that people are far more likely to read the post than the comments and comment responses.
Also, if you don’t include objections and counter-arguments, you run the risk that someone will read your post, think of an objection, fail to think of a counter-objection, and dismiss the idea, where s/he might have accepted it if you had included more.
True. Perhaps some posts should include separate “anticipation of objections” parts. These could be skipped by people who prefer terse posts but would have to be read by anyone who wishes to comment.
Or we could do the opposite pattern and try to have a high-level takeaway/summary that in theory could get you 80% of the value of reading the post in 20% of the time.
We could call it an “abstract”! Or a “tl;dr”, for the younger generation.
Most LW posts don’t have them, and they typically aren’t comprehensive enough to get you 80% of the value of reading the entire post.
They can tell you whether a post is worth reading, and give the writer a target to aim at.
Yes, Unfortunately, to assume a querulous reader is largely correct.
Nit-picky objections can be addressed in footnotes[1]. And you can address legitimate qualifications and caveats by first giving a broad, intuitive overview of what you believe, while cautioning that the position you can actually stand for is more complicated than that.
[1] Since some folks may be unfamiliar with them, here’s how you actually add a footnote to your comment or post.
No.
get out of my head
“Anyone can build a bridge. It takes an engineer to barely build a bridge.”
I heard this on a train, I’ve never been able to find the source, and I use it to explain the difference between solving a problem and elegantly solving a problem. Another perspective (attributed to R. Buckminster Fuller but I’m not confident about that attribution):
“When I’m working on a problem, I never think about beauty. I think only how to solve the problem. But when I have finished, if the solution is not beautiful, I know it is wrong.”
It seems like you have several separate things in mind: readability, information density, arguments masquerading as true causes of beliefs, trustworthiness of “experienced rationalists,” and the value of the “main point” vs disclaimers and qualification. Do you have an example post in mind, and specific suggestions for an improved version? I’m not sure if I’m about to respond to you, or just ramble. I understand that it’s hard to call someone out without being mean, but these meta discussions seem to go nowhere without specific examples. (That said, the rest of my comment is not any better in this regard. Also, feel free to use any of my posts or comments for target practice.)
I agree that LW falls short on readability. Most people, LW posters included, are not good writers to begin with. Conciseness is one absent virtue among many. On the other hand, extremely information-dense texts can also be unreadable. Also, conversational, forceful, and polemical styles—which are easier and more entertaining to read than academic styles heavy with caveats—accompany lower epistemic standards. It’s possible to be highly readable, informative, and rigorously correct, but it’s hard. Littering your claims with “I think”s and “probably”s is a poor solution. Even if your main worry is coming off as too authoritative, such ugly filler qualifying language can be replaced with specific qualifications, possibly in a preface or footnote.
Your “murder is wrong” example is a poorly-constructed sentence, sure, for reasons beyond the above. But the details and qualifications are the real content of that statement. I don’t think that’s because “murder is wrong” is an uncharacteristically content-free claim. For any basic principle or sweeping generalization, there will be cases where it obviously works, cases where it obviously doesn’t, and the real information is in where you draw the line. (And that statement applies to itself. I may need to elaborate in a separate post.)
I already see over-reliance on simple/abstract/principled arguments and beliefs as a weakness of LW discourse. This is a shame, because people here should have a huge advantage in terms of consequentialist reasoning and the ability to recognize and discuss tradeoffs without knee-jerk responses.
Regarding your other points, I agree that people often present arguments which do not include the true causes of their beliefs, and that this is bad. I also (relatedly) have enough confidence in very few people here such that a mere statement of their beliefs would be informative.
You’re right, there’s multiple related points to be made here. So, let me try to clarify the issue a bit.
If I write a post with specific examples, people will get too hung up on the examples and the main point is lost.
if I don’t post examples, people will comment that I am not specific enough.
If I state very clearly and succinctly what I mean, there’s the risk that I come across as arrogant.
If I quantify everything with disclaimers, the post becomes less readable and people won’t know what I think.
If I don’t anticipate every possible counter-argument, people will assume I haven’t thought of those arguments.
If I do anticipate and counter every counter-argument, the post becomes less readable again.
It’s a damned if you do, damned if you don’t situation. It really is hard, as you say. I originally included in the text that I was thinking about the meta-ethics sequence specifically: I feel that by attempting to include every possible concern, every possible argument and counter-argument, the sequence became way too long and unreadable. At no point does Yudkowsky clearly and succinctly state what his position is. As a result, I think the point of the sequence was lost on people.
Yet there’s a reason for this. Eliezer isn’t stupid: I think he wrote the sequence that way precisely because he feared that if he did state clearly what he meant right of the bat, people would instantly reject it based on trivial arguments. However, I think that the solution is for the reader-base to appreciate frankness and terseness more, and demand argumentation and qualification less. Not to demand that people like Yudkowsky write perfect posts that cannot possibly be misinterpreted and cover everything without being too long.
We post our beliefs hoping to convince other people. Even if we’re not certain of a belief, we generally intend for the readers to approach our own degree of confidence.
We give the best arguments we know for the beliefs (and if we’re good rationalists and are also honest and confident, we also post the best arguments we know against them.) Sometimes these aren’t the true reasons we hold those beliefs, because we’re not luminous or introspective enough. But they are still arguments that can be evaluated and they are legitimate, admissible evidence that people should indeed update on.
If we didn’t give any evidence but just stated our beliefs, then the only reason anyone would be convinced by them would be that they trusted our strength as rationalists. And such trust is generally misplaced, precisely because as you point out, even when we try to honestly analyze why we believe something, we are often mistaken, biased, etc.
If you also consider dishonest, manipulative, or just plainly irrational actors, then the mere fact they claim to believe something is very hard to update on correctly. Whereas the justifications they give for the belief can still be evaluated and provide at least some evidence.
I like terseness also (so does Hanson).
But I believe in letting a thousand flowers bloom. Some folks write a lot, because they are at their best this way. Who am I to tell them they are wrong?
Even here, I only trust about three people’s epistemic standards enough to follow up on two-sentence summaries of their opinions. More if they’re domain experts. Less if the subject’s controversial.
Opinions unsupported by argument are of no value to me. Why are they to you? Are they useful for something? Or do you value them terminally?
Oops, wrong reply
I think a fair portion of Lesswrong actually has complex reasons for holding the beliefs that they are holding.
On most interesting question I have a position that evolved over years through being exposed to different arguments. If I write a full version of why I believe what I believe that’s going to be long. I rather focus on making those points that might be important for the audience with whom I’m interacting.
Journalism has been infected with a style of stream of consciousness random anecdotes. Apparently many go for that. I don’t. I read a few paragraphs, start to get agitated, and recall “When you talk like this I can’t help but wonder—is there a point?” Verbal fog of this kind is a good sign of a poorly thought out idea, in my experience, or no particular idea at all. But if you blow enough smoke, you can obscure what’s happening until the end of the article. For some, that’s ideal—passing the time and “justifying” a little check next to “I read an article; I’m a smart and informed person”.
If the goal is communication, terse is good. The more concise and streamlined you can make your message, the easier it is to see the relationship between the parts and grok it.
What helps on longer arguments is to give the reader a roadmap of where you’re going, why you’re going, and how you plan to get there. That roadmap gives the reader motivation to read, and sign posts so he can know keep track of far he’s come along the trip.
Particularly on the internet, you’d better be prepared to motivate a different type of reader with a payoff, and keep him motivated by showing him along the way that you’re continuing to make progress toward that payoff for him.
Some people want to pass the time with a soothing ride through the fog, but I think this list has fewer of those types and more of the latter types. We want a payoff, and will move on when we don’t see one coming.
Unless your idea is very obvious, most people will absorb it better if you offer examples.
Personally, I like long posts because I need time to an idea things through, and do that more efficiently while I am also reading about that idea, instead of reading it through then finishing and becoming vulnerable to the distractions of the internet again.
It took you more than a sentence or two to explain why you like terseness.
A) It would hardly make sense to assume the correctness of my conclusion in my defence of it.
B) The “sentence or two” was in reference to a “list of things rational people believe”, which this post is not.
C) “A sentence or two” should not be taken to literally mean “Either 1 or 2 sentences”. It’s a colloquialism.
A) reaching clear success within its constraints would be a good sign
terseness is good because it’s a lot harder to make a simple idea or belief needlessly complex if you’re terse, but it’s also hard to use it to explain complex things.
I strongly agree with this post, people on this site talk way, way more than they need to in order to make the point.
Clarity is of course desirable. Shortness is not always helpful.
Here is an example of a long post that requires a good deal of reader perseverance to arrive at its main point. To wit, CDC obesity studies since the mid-20th century underwent a change in demographic sampling partway through (with more blacks and Hispanics sampled), resulting in a likely overstatement of obesity trend statistics.
The post title gives a hint, but the article would have been improved by indicating where it was headed much earlier on.
In contrast, this post delivers its message regarding the interpretation of “accurate more than 90% of the time” (including definitions of sensitivity and specificity) in a straightforward manner. I wouldn’t describe the post as terse, but I give it high marks for content/length.
I had already noticed I needed to adjust in this direction. I’m going to try being more concise and see how it goes.
Specifically, 50 words or less for 50 comments. (#50)
The signature lines you’re writing under your posts are confusing and caused me to follow your post history to understand what was going on, arriving here, so that I can now tell you adding a block of extra text to all your posts in an effort to be terse seems like a lost purpose.
Thanks for the feedback. I’ll keep track of my 50 comments more unobtrusively. (Comment #47.)
How do you decide whether to count down or to count up?
I preferred to count down since I would like to keep track of how many comments remain until I’ve successfully met my commitment. If I had just wanted to accumulate an unspecified number, I would have counted up.
…any particular reason why you asked?
Just because you are counting up doesn’t mean the limit is unspecified. Indeed, your original version explicitly specified the limit in every comment.
I asked because it looked very odd to me, especially counting down using # signs. I asked in the present tense because I thought you might have a general rule, such as counting down for commitments. Here is an analogous situation: what if you are managing a crowd on a trip, so you count them at the beginning, and again at some checkpoint, to make sure you haven’t lost them. Do you count down?