[My comment kind-of seems insanely long for this topic, but āwhat to call a postā is a decision Iāll probably face hundreds of times in future, so it seems worth thinking about it more.]
I expect we disagree, though Iām not sure how much, and Iād be interested in trying to find the crux and seeing whether one of us will end up changing our minds.
First, Iāll gesture at my broader views on related topics by adapting a comment I wrote in a doc. Though this comment isnāt specifically on numbered list titles, and Iād say different things about numbered list titles specifically. (This comment was prompted by someone suggesting the doc be given a more āfunā title.)
I agree that āfunā titles get more readership, but I think that this is bad and should not be exploited, and I think that itās good that the Forum already has more of a mild norm against exploiting this.
To expand: I think our goal should be for people to (1) read whatever is most useful to them, or perhaps whatever they can provide the most value to (via comments, contacting the author to share thoughts, etc.), and also for people to (2) spend the appropriate amount of time on reading vs other activities.
For that goal, it could help to be more attention-grabbing/āfun, if either (a) we have good reason to think our thing is whatās best for them to read but that theyāll fail to appreciate that by default, or (b) people are reading things like Forum posts less than they should (e.g., they get bored and go play video games).
But I think we should be skeptical about (a), just as weād be skeptical of other people saying it. And I donāt think (b) is as big an issue as the choice of what to read (plus some people may read the Forum more than they should, though Iād guess thatās less common than reading it less than they should, even among EAs).
So I think we should aim for titles to basically just make the scope and purpose clear and not be especially boring or long.
If we can get more attention-grabbing-ness without sacrificing clarity of purpose and scope, that could sometimes be worthwhile, but Iām wary even of that. E.g., for this reason, I called a post āSome history topics it might be very valuable to investigateā rather than ā10 history topics it might be very valuable to investigateā, even though it literally did happen to be 10.
And if weād have to sacrifice clarity, then I think thatās very rarely worthwhile.
Relatedly, I think the Forumās mild norm against exploiting attention-grabbing-ness is a good thing and should be maintained. Even if we believed (a) and it was true, probably a bunch of other people would believe similar things about themselves. And weāre all better off if we just make it as easy as possible for readers to work out what itād make sense for them to read, rather than trying to entice them or grab their attention.
I think itās sort-of like weāre in a community-level prisonerās dilemma, and currently weāre cooperating, and we should stick with that.
(Here I partly have in mindScott Alexanderās writing on asymmetric weapons:āLogical debate has one advantage over narrative, rhetoric, and violence: itās an asymmetric weapon. That is, itās a weapon which is stronger in the hands of the good guys than in the hands of the bad guys.ā)
And in this case, I think [personās] suggestion would indeed sacrifice clarity - [reason why the āfunā title would make the scope and purpose of the post less immediately clear than other titles].
---
So I guess there were really three key downsides I was pointing to in that comment:
Optimising partly for attention-grabbing-ness in titles might in expectation worsen readersā choices about what to read and how long to spend reading things in a given venue, because it sometimes trades of against being clear about the purpose and scope of a piece
Optimising partly for attention-grabbing-ness in titles might in expectation worsen readersā choices about what to read and how long to spend reading things in a given venue, because it introduces an āattractorā that isnāt necessarily at all correlated with how worth-reading something is for a given person
Optimising partly for attention-grabbing-ness is kind-of like defecting on a prisonerās dilemma, and may lead other people to do the same, which could be bad even if it really is true that what āyouā are writing is especially worthy of readersā attention and āyourā title choice would still be clear
When I wrote the above comment, it was in a situation where I think all 3 of those downsides applied. Just swapping āSome [blah]ā for ā10 [blah]ā doesnāt face downside 1, which makes it much better. So Iām probably actually more ok with numbered list titles than some other types of āfunā titles (e.g., off the top of my head, calling this post āHistory, the long-term, and youā or āHistory! What is it good for?ā).
But I think thereās also a fourth possible downside:
4. Using a numbered list title for a post may also make you, or other people, more likely to write posts suited to numbered list formats, or squeeze other posts into that format (and maybe into neat or not-too-high numbers). This would happen if numbered list titles increased attention in expectation, and so people were incentivised to use them, and perhaps even would simply struggle to ācompeteā for attention if they donāt use them in an environment where everyone else is.
So I think, when it comes to numbered list titles for posts that do fit that mould, the possible downsides are 2, 3, and 4, and how strong those downsides are depends to a significant extent (though not entirely) on the extent to which numbered list titles grab attention in a way that isnāt correlated with whatās useful for the reader or where the reader can provide value.
So, if you basically agree with the above framings in theory, maybe the crux between you and I is just the empirical question of the extent to which numbered list titles do that. (Related to your statement ā(3) I donāt find it very plausible that a straightforward declaration of the size of a list tricks people into reading things they otherwise ought not to have (while I agree for phrases like āNumber 5 will SHOCK you,ā or outrage bait)ā. Though Iām not sure I like the word ātricksā.)
Do you think thatās the crux? If so, maybe google could quickly reveal answers to that empirical question?
---
Also, do you think you agree with my views when it comes to āfunā titles that are unclear as to purpose and scope (e.g., āHistory, the long-term future, and youā), separate from the question of numbered list titles?
---
(Also, I should note that all of this is specifically in the context of the EA ForumāIām more open to āattention-grabbingā strategies in other contexts. I could elaborate if readers are interested, but Linch and I already discussed that separately.)
Hmm, I have two somewhat different claims first, before engaging with the structure of your argument.
The numbered list format is actually just a great communicative technology, both in general and especially well-adapted to the internet age.
In a nonadversarial context, honest signaling of traits is all-else-equal very good.
Itās possible you already agree with these points, so I will use them as assumptions and wonāt defend them further here unless requested.
extent to which numbered list titles grab attention in a way that isnāt correlated with whatās useful for the reader or where the reader can provide value
I think this is a large difference between us, but not by itself the ultimate crux. I think thereās another consideration:
extent to which numbered list titles are superior to other titles in their ability to provide value to the user.
I agree that if the costs of numbered list titles is high enough, we probably shouldnāt do them. However, I think between us we diverge on the differences in our evaluation of the benefits of numbered list titles as well.
Broadly, I think numbered list titles are useful honest signaling^ to quickly tell readers whether to engage with a post (and also helpfully whether to stop reading a post, after they read the first point in a list).
For another example:
Using a numbered list title for a post may also make you, or other people, more likely to write posts suited to numbered list formats, or squeeze other posts into that format
To me (to the extent that your observation is correct), this is evidence that the marketplace of ideas desire more articles well-suited to numbered list formats, so itās all-else-equal evidence that people ought to write more numbered lists. (Analogously, if it turns out a lot of EA people like listening to podcasts, I think this is evidence that more EAs should make podcasts, even if I personally hate podcasts). So something that to you is a cost I mostly think of as a benefit. (I donāt think the EA marketplace of ideas is always accurate, for example it rewards culture war posts more than I perhaps think is fair).
Ultimately, if it turns out that numbered list titles attract people more to reading posts that they otherwise wouldnāt have, and this is a bad choice of their time, I agree that this is usually a bad idea.
I think my main heuristic for preferring to title numbered lists when the article is essentially a numbered list to begin with is that honest signaling is usually good, both in general and specific to the question of EA articles . There are exceptions of course^^, however Iām not convinced that one of the core exceptions applies here.
^The honest part here is relevant to me, which is why I originally contrasted with the phrase ātricked.ā
^^ for example if the signaling is very costly (broadly, getting a PhD or becoming a chess grandmaster just to signal intelligence and conscientiousness, narrowly, spending a bajillion hours to make a post look closer to an academic preprint even when academia is not your target audience), when the signaling causes fixation on a trait we donāt care about (broadly, itās bad to include glamor pictures unnecessarily online to take advantage of physical attractiveness; narrowly, itās prima facie bad to include unhelpful Latin phrases to signal erudition)
Hmm. It seems like weāve indeed identified the two cruxes.
Regarding the benefits: I donāt see why numbered list titles would help a reader make good decisions about whether to engage with a post? In particular, given that they could in any case use the title (whatever its form), the summary-type thing (which ideally there would be) and/āor the first paragraph (if thatās different), the word count /ā scrolling to see how long it seems, and the karma and comments?
Regarding the extent to which numbered list titles grab attention in a way that isnāt correlated with whatās useful for the reader or where the reader can provide value: Maybe at some point I should look for empirical evidence, or at least better theorising, regarding this. Currently I think we just have different intuitions/āanecdata.
In particular, given that they could in any case use the title (whatever its form), the summary-type thing (which ideally there would be) and/āor the first paragraph (if thatās different), the word count /ā scrolling to see how long it seems, and the karma and comments?
Maybe Iām missing something, but I feel like this is an always-general argument against all informative title names?
Maybe at some point I should look for empirical evidence, or at least better theorising, regarding this. Currently I think we just have different intuitions/āanecdata.
I agree that we have different intuitions and empirical data may help resolve this.
Maybe Iām missing something, but I feel like this is an always-general argument against all informative title names?
I donāt think soāI think itās quite clear how itās easier for a reader to make good choices about whether to read this post if itās called āSome history topics it might be very valuable to investigateā than if it was called (for example) āTopicsā or āHistory stuffā or āWhat you can do with historyā. But I just donāt immediately see why changing it from the current title to ā10 history topics it might be very valuable to investigateā would help the reader make good choices?
It seems like whether itās about history and whether itās research topics is useful info, but whether itās 3 or 10 or 20 isnāt very useful, especially given that I probably couldāve included roughly the same content under 3 topics or split it up into 20.
And then the word count /ā scrolling is relevant because, if the consideration is āhow long will this take me?ā, then word count /ā scrolling seems to address that better than reading one topic and multiplying by the stated number of topics. (The latter requires reading a topic before deciding, and the topics may actually differ in length.)
(Of course, there may be a reason Iām missing; I wouldnāt be that surprised if you said one sentence and then I went āOh yeah, fair point, I shouldāve thought of that.ā)
[My comment kind-of seems insanely long for this topic, but āwhat to call a postā is a decision Iāll probably face hundreds of times in future, so it seems worth thinking about it more.]
I expect we disagree, though Iām not sure how much, and Iād be interested in trying to find the crux and seeing whether one of us will end up changing our minds.
First, Iāll gesture at my broader views on related topics by adapting a comment I wrote in a doc. Though this comment isnāt specifically on numbered list titles, and Iād say different things about numbered list titles specifically. (This comment was prompted by someone suggesting the doc be given a more āfunā title.)
---
So I guess there were really three key downsides I was pointing to in that comment:
Optimising partly for attention-grabbing-ness in titles might in expectation worsen readersā choices about what to read and how long to spend reading things in a given venue, because it sometimes trades of against being clear about the purpose and scope of a piece
Optimising partly for attention-grabbing-ness in titles might in expectation worsen readersā choices about what to read and how long to spend reading things in a given venue, because it introduces an āattractorā that isnāt necessarily at all correlated with how worth-reading something is for a given person
Optimising partly for attention-grabbing-ness is kind-of like defecting on a prisonerās dilemma, and may lead other people to do the same, which could be bad even if it really is true that what āyouā are writing is especially worthy of readersā attention and āyourā title choice would still be clear
When I wrote the above comment, it was in a situation where I think all 3 of those downsides applied. Just swapping āSome [blah]ā for ā10 [blah]ā doesnāt face downside 1, which makes it much better. So Iām probably actually more ok with numbered list titles than some other types of āfunā titles (e.g., off the top of my head, calling this post āHistory, the long-term, and youā or āHistory! What is it good for?ā).
But I think thereās also a fourth possible downside:
4. Using a numbered list title for a post may also make you, or other people, more likely to write posts suited to numbered list formats, or squeeze other posts into that format (and maybe into neat or not-too-high numbers). This would happen if numbered list titles increased attention in expectation, and so people were incentivised to use them, and perhaps even would simply struggle to ācompeteā for attention if they donāt use them in an environment where everyone else is.
So I think, when it comes to numbered list titles for posts that do fit that mould, the possible downsides are 2, 3, and 4, and how strong those downsides are depends to a significant extent (though not entirely) on the extent to which numbered list titles grab attention in a way that isnāt correlated with whatās useful for the reader or where the reader can provide value.
So, if you basically agree with the above framings in theory, maybe the crux between you and I is just the empirical question of the extent to which numbered list titles do that. (Related to your statement ā(3) I donāt find it very plausible that a straightforward declaration of the size of a list tricks people into reading things they otherwise ought not to have (while I agree for phrases like āNumber 5 will SHOCK you,ā or outrage bait)ā. Though Iām not sure I like the word ātricksā.)
Do you think thatās the crux? If so, maybe google could quickly reveal answers to that empirical question?
---
Also, do you think you agree with my views when it comes to āfunā titles that are unclear as to purpose and scope (e.g., āHistory, the long-term future, and youā), separate from the question of numbered list titles?
---
(Also, I should note that all of this is specifically in the context of the EA ForumāIām more open to āattention-grabbingā strategies in other contexts. I could elaborate if readers are interested, but Linch and I already discussed that separately.)
Hmm, I have two somewhat different claims first, before engaging with the structure of your argument.
The numbered list format is actually just a great communicative technology, both in general and especially well-adapted to the internet age.
In a nonadversarial context, honest signaling of traits is all-else-equal very good.
Itās possible you already agree with these points, so I will use them as assumptions and wonāt defend them further here unless requested.
I think this is a large difference between us, but not by itself the ultimate crux. I think thereās another consideration:
I agree that if the costs of numbered list titles is high enough, we probably shouldnāt do them. However, I think between us we diverge on the differences in our evaluation of the benefits of numbered list titles as well.
Broadly, I think numbered list titles are useful honest signaling^ to quickly tell readers whether to engage with a post (and also helpfully whether to stop reading a post, after they read the first point in a list).
For another example:
To me (to the extent that your observation is correct), this is evidence that the marketplace of ideas desire more articles well-suited to numbered list formats, so itās all-else-equal evidence that people ought to write more numbered lists. (Analogously, if it turns out a lot of EA people like listening to podcasts, I think this is evidence that more EAs should make podcasts, even if I personally hate podcasts). So something that to you is a cost I mostly think of as a benefit. (I donāt think the EA marketplace of ideas is always accurate, for example it rewards culture war posts more than I perhaps think is fair).
Ultimately, if it turns out that numbered list titles attract people more to reading posts that they otherwise wouldnāt have, and this is a bad choice of their time, I agree that this is usually a bad idea.
I think my main heuristic for preferring to title numbered lists when the article is essentially a numbered list to begin with is that honest signaling is usually good, both in general and specific to the question of EA articles . There are exceptions of course^^, however Iām not convinced that one of the core exceptions applies here.
^The honest part here is relevant to me, which is why I originally contrasted with the phrase ātricked.ā
^^ for example if the signaling is very costly (broadly, getting a PhD or becoming a chess grandmaster just to signal intelligence and conscientiousness, narrowly, spending a bajillion hours to make a post look closer to an academic preprint even when academia is not your target audience), when the signaling causes fixation on a trait we donāt care about (broadly, itās bad to include glamor pictures unnecessarily online to take advantage of physical attractiveness; narrowly, itās prima facie bad to include unhelpful Latin phrases to signal erudition)
Hmm. It seems like weāve indeed identified the two cruxes.
Regarding the benefits: I donāt see why numbered list titles would help a reader make good decisions about whether to engage with a post? In particular, given that they could in any case use the title (whatever its form), the summary-type thing (which ideally there would be) and/āor the first paragraph (if thatās different), the word count /ā scrolling to see how long it seems, and the karma and comments?
Regarding the extent to which numbered list titles grab attention in a way that isnāt correlated with whatās useful for the reader or where the reader can provide value: Maybe at some point I should look for empirical evidence, or at least better theorising, regarding this. Currently I think we just have different intuitions/āanecdata.
Maybe Iām missing something, but I feel like this is an always-general argument against all informative title names?
I agree that we have different intuitions and empirical data may help resolve this.
I donāt think soāI think itās quite clear how itās easier for a reader to make good choices about whether to read this post if itās called āSome history topics it might be very valuable to investigateā than if it was called (for example) āTopicsā or āHistory stuffā or āWhat you can do with historyā. But I just donāt immediately see why changing it from the current title to ā10 history topics it might be very valuable to investigateā would help the reader make good choices?
It seems like whether itās about history and whether itās research topics is useful info, but whether itās 3 or 10 or 20 isnāt very useful, especially given that I probably couldāve included roughly the same content under 3 topics or split it up into 20.
And then the word count /ā scrolling is relevant because, if the consideration is āhow long will this take me?ā, then word count /ā scrolling seems to address that better than reading one topic and multiplying by the stated number of topics. (The latter requires reading a topic before deciding, and the topics may actually differ in length.)
(Of course, there may be a reason Iām missing; I wouldnāt be that surprised if you said one sentence and then I went āOh yeah, fair point, I shouldāve thought of that.ā)