The commenting restrictions on LessWrong seem bad
I wrote a few controversial articles on LessWrong recently that got downvoted. Now, as a consequence, I can only leave one comment every few days. This makes it totally impossible to participate in various ongoing debates, or even provide replies to the comments that people have made on my controversial post. I can’t even comment on objections to my upvoted posts. This seems like a pretty bad rule—those who express controversial views that many on LW don’t like shouldn’t be stymied from efficiently communicating. A better rule would probably be just dropping the posting limit entirely.
I think being unable to reply to comments on your own posts is very likely a mistake and we should change that. (Possibly if the conditions under which we think that was warranted, we should issue a ban.)
”I’m downvoted because I’m controversial” is a go-to stance for people getting downvoted (and resultantly rate-limited), though in my experience the issue is quality rather than controversy (or rather both in combination).
Overall though, we’ve been thinking about the rate limit system and its effects. I think there are likely bad effects even if it’s successfully in some case reducing low quality stuff.
That’s usually true but IMO in this case the heavy downvoting was pretty tribalistic in nature. Something about the subject matter makes people think that opposing views are clearly crazy, and they don’t bother trying to understand(admittedly omnizoid’s posts suffered from this as well)
I was about to write a comment to the effect that there should clearly be an exception for commenting on your own posts (and, indeed, anyone who can’t even be allowed to comment on their own posts should just be banned), so… yeah, strongly agreed that this particular thing should be fixed!
EDIT: Never mind, I see this has already been fixed. Excellent!
? (EDIT: I mean that the quoted sentence is currently missing some word and is hence incomprehensible.)
I agree that people should probably be able to reply to comments on their own posts. However, if enabling this is a non-trivial amount of work, I suspect the LW team’s time would be better spent elsewhere.
I base this on the presumptions that 1) there aren’t too many people this policy would help (dozens? single-digits?), 2) these people wouldn’t bring much value to the community, and 3) such a policy is unlikely to be deterring people we’d otherwise want from joining and contributing to the community.
Both views seem symmetric to me:
They were downvoted because they were controversial (and I agree with it / like it).
They were downvoted because they were low-quality (and I disagree with it / dislike it).
Because I can sympathize with both views here, I think we should consider remaining agnostic to which is actually the case.
It seems like the major crux here is whether we think that debates over claim and counter-claim (basically, other cruxes) are likely to be useful or likely to cause harm. It seems from talking to the mods here and reading a few of their comments on this topic that they tend to learn towards them being harmful on average and thus need to be pushed down a bit.
Since omnizoid’s issue is not merely over issues of quality, but both over quality as well as being counter-claims to specific claims that have been dominant on LessWrong for some time.
The most agnostic side of the “top-level” crux that I mentioned above seems to point towards favoring agnosticism and furthermore that if we predict debates to be more fruitful than not, then one needn’t be too worried even if one is sure that one side of another crux is truly the lower-quality side of it.
This is, as far as I can tell, totally false. There is a very different claim one could make which at least more accurately represents my opinion, i.e. see this comment by John Wentworth (who is not a mod).
Most of your comment seems to be an appeal to modest epistemology. We can in fact do better than total agnosticism about whether some arguments are productive or not, and worth having more or less of on the margin.
Note that the more you believe that your commenters can tell whether some arguments are productive or not, and worth having more or less of on the margin, the less you should worry as mods about preventing or promoting such arguments (altho you still might want to put them near the top or bottom of pages for attention-management reasons)
My actual belief is that commenters can (mostly) totally tell which arguments are productive… but… it’s hard to not end up having those unproductive arguments anyway, and the site gets worse.
Raemon’s comment below indicates mostly what I meant by:
Furthermore, I think the mods’ stance on this is based primarily on Yudkowsky’s piece here. I think the relevant portion of that piece is this (emphases mine):
So, it seems to me that the relevant issues are the following. Being more tolerant of lower-quality discussion will cause:
Higher-quality members’ efforts being directed toward less fruitful endeavors than they would otherwise be.
Higher-quality existing members to leave the community.
Higher-quality potential members who would otherwise have joined the community, not to.
My previous comment primarily refers to the notion of the first bullet-point in this list. But “harmful on average” also means all three.
The issue I have most concern with is the belief that lower-quality members are capable of dominating the environment over higher-quality ones, with all-else-being-equal, and all members having roughly the same rights to interact with one another as they see fit.
This mimics a conversation I was having with someone else recently about Musk’s Twitter / X. They have different beliefs than I do about what happens when you try to implement a system that is inspired by Musk’s ideology. But I encountered an obstacle in this conversation: I said I have always liked using it [Twitter / X], and it also seems to be slightly more enjoyable to use post-acquisition. He said he did not really enjoy using it, and also that it seems to be less enjoyable to use post-acquisition. Unfortunately, if it comes down to a matter of pure preferences like this, than I am not sure how one ought to proceed with such a debate.
However, there is an empirical observation that one can make comparing environments that use voting systems or rank-based attention mechanisms: It should appear to one as though units of work that feel like more or better effort was applied to create them correlate with higher approval and lower disapproval. If this is not the case, then it is much harder to actually utilize feedback to improve one’s own output incrementally. [1]
On LessWrong, that seems to me to be less the case than it does on Twitter / X. Karma does not seem correlated to my perceptions about my own work quality, whereas impressions and likes on Twitter / X do seem correlated. But this is only one person’s observation, of course. Nonetheless I think it should be treated as useful data.
That being said, it may be that the intention of the voting system matters: Upvotes / downvotes here mean “I want to see more of / I want to see less of” respectively. They aren’t explicitly used to provide helpful feedback, and that may be why they seem uncorrelated with useful signal.
We don’t currently have karma and agree/disagree voting broken out separately on posts (like we do on comments). But I’m pretty sure that if we did, your posts would still be heavily downvoted. The problem is not that people disagree with you; “controversial views that many on LW don’t like” isn’t the central issue. The problem is that you’re presenting arguments which have been repeatedly floated for years and which most long-time LWers have long since dismissed as dumb. This is being correctly rate-limited, for exactly the same reasons that the site would downvote (and then rate-limit) e.g. someone who came along arguing that we can’t dismiss the existence of God because epistemic modesty, or someone who came along arguing that Searle’s Chinese room argument proves AI will never be an X-risk, or some similarly stupid (but not-completely-dead-outside-LW) argument.
Oh, come on, it’s clear that the Yudkowsky post was downvoted because it was bashing Yudkowsky and not because the arguments were dismissed as “dumb.”
It wouldn’t have mattered to me whose name was in the title of that post, the strong-downvote button floated nearer to me just from reading the rest of the title.
From reading omnizoid’s blog, he seems overconfident in all his opinions. Even when changing them, the new opinion is always a revelation of truth, relegating his previously confident opinion to the follies of youth, and the person who bumped him into the change is always brilliant.
“It wouldn’t have mattered to me whose name was in the title of that post, the strong-downvote button floated nearer to me just from reading the rest of the title.”
I think this is right from an individual user perspective, but misses part of the dynamic. My impression from reading lesswrong posts is that something like that post, had the topic been different, maybe “Jesus was often incredibly wrong about stuff”, would have been ignored by many people. It would maybe have had between zero and a dozen karma and clearly not been clicked on by many people.
But that post, in some sense, was more successful than ones that are ignored—it managed to get people to read it (which is a necessary first step of communicating anything). That it has evidently failed in the second step (persuading people) is clear from the votes.
In a sense maybe this is the system working as intended: stuff that people just ignore doesn’t need downvoting because it doesn’t waste much communication bandwidth. Where as stuff that catches attention then disapoints is where the algorithm can maybe do people a favour with downvote data. But the way that system feeds into the users posting rights seems a little weird.
There are plenty of people on LessWrong who are overconfident in all their opionions (or maybe write as if they are, as a misguided rhetorical choice?). It is probably a selection effect of people who appreciate the sequences—whatever you think of his accuracy record, EY definitely writes as if he’s always very confident in his conclusions.
Whatever the reason, (rhetorical) overconfidence is most often seen here as a venial sin, as long as you bring decently-reasoned arguments and are willing to change your mind in response to other’s. Maybe it’s not your case, but I’m sure many would have been lighter with their downvotes had the topic been another one—just a few people strong downvoting instead of simple downvoting can change the karma balance quite a bit
Yeah, I agree I have lots of views that LessWrongers find dumb. My claim is just that it’s bad when those views are hard to communicate on account of the way LW is set up.
I think it’s not just the views but also (mostly?) the way you write them.
This is hindsight, but next time instead of writing “I think Eliezer is often wrong about X, Y, Z” perhaps you should first write three independent articles “my opinion on X”, “my opinion on Y”, my opinion on Z”, and then one of two things will happen—if people agree with you on X, Y, Z, then it makes sense to write the article “I think Eliezer is often wrong” and use these three articles as evidence… or if people disagree with you on X, Y, Z, then it doesn’t really make sense to argue to that audience that Eliezer is wrong about that, it they clearly think that he actually is right about X, Y, Z. If you want to win this battle, you must first win the battles about X, Y, Z individually.
(Shortly, don’t argue two controversial things at the same time. Either make the article about X, Y, Z, or about Eliezer’s overconfidence and fallibility. An argument “Eliezer is wrong because he says things you agree with” will not get a lot of support.)
Alternatively, it can happen that people will disagree with you on X, Y, but agree on Z. In that case you can still make an argument for “Eliezer is sometimes wrong” and only use the discussion on Z as an example.
As shminux describes well, it’s possible to write about controversial views in a way that doesn’t get downvoted into nirvana. To do that, you actually have to think about how to write well.
The rate limits, limits the quantity but that allows you to spend more time to get the quality right. If you are writing in the style you are writing you aren’t efficiently communicating in the first place. That would require to think a lot more about what the cruxes actually are.
I don’t think people recognize when they’re in an echo chamber. You can imagine a Trump website downvoting all of the Biden followers and coming up with some ridiculous logic like, “And into the garden walks a fool.”
The current system was designed to silence the critics of Yudkowski’s et al’s worldview as it relates to the end of the world. Rather than fully censor critics (probably their actual goal) they have to at least feign objectivity and wait until someone walks into the echo chamber garden and then banish them as “fools”.
As someone with significant understanding of ML who previously disagreed with yudkowsky but have come to partially agree with him on specific points recently due to studying which formalisms apply to empirical results when, and who may be contributing to downvoting of people who have what I feel are bad takes, some thoughts about the pattern of when I downvote/when others downvote:
yeah, my understanding of social network dynamics does imply people often don’t notice echo chambers. agree.
politics example is a great demonstration of this.
But I think in both the politics example and lesswrong’s case, the system doesn’t get explicitly designed for that end, in the sense of people bringing it into a written verbal goal and then doing coherent reasoning to achieve it; instead, it’s an unexamined pressure. in fact, lesswrong is explicit-reasoning-level intended to be welcoming to people who strongly disagree and can be precise and step-by-step about why. However,
I do feel that there’s an unexamined pressure reducing the degree to which tutorial writing is created and indexed to show new folks exactly how to communicate a claim in a way lesswrong community voting standards find upvoteworthy-despite-disagreeworthy. Because there is an explicit intention to not fall to this implicit pressure, I suspect we’re doing better here than many other places that have implicit pressure to bubble up, but of course having lots of people with similar opinions voting will create an implicit bubble pressure.
I don’t think the adversarial agency you’re imagining is quite how the failure works in full detail, but because it implicitly serves to implement a somewhat similar outcome, then in adversarial politics mode, I can see how that wouldn’t seem to matter much. Compare peer review in science: it has extremely high standards, and does serve to make science tend towards an echo chamber somewhat, but because it is fairly precisely specified what it takes to get through peer review with a claim everyone finds shocking—it takes a well argued, precisely evidenced case—it is expected that peer review serves as a filter that preserves scientific quality. (though it is quite ambiguous whether that’s actually true, so you might be able to make the same arguments about peer review! perhaps the only way science actually advances a shared understanding is enough time passing that people can build on what works and the attempts that don’t work can be shown to be promising-looking-but-actually-useless; in which case peer review isn’t actually helping at all. but I do personally think step-by-step validity of argumentation is in fact a big deal for determining whether your claim will stand the test of time ahead of time.)
As someone who gets rate-limited due to downvoted comments occasionally, I can see a rather strong correlation between the style/tone and the response. It is very much possible to express controversial, contrarian or even outright silly views without being downvoted. The rule of thumb is to be respectful and charitable to your opponents and, well, read the room. The more your view diverges from the local mainstream, the better, tighter and clearer your argument must be.
It is standard practice when writing scientific papers to first demonstrate intimate familiarity with the views you argue against later. This is what you would want from someone else writing a refutation of your argument, right? So give them the same courtesy. You can cut corners if your post or comment is close to the mainstream, but it is not a great habit. If you want to learn how to do it right, read ACX, Scott is a master of this art.
it is a common pitfall to blame the society/subculture for the negative reaction you get. It is almost never a constructive way to proceed.
If you are in doubt as to how your post would be received, and you want to get through to more people, consider reaching out to someone familiar with this site to review your draft. Or to anyone, really. The bar for writing a post or a comment with non-negative expected karma is pretty low.
Regarding this, I like to quote Scott’s rules for commenting on his sites:
Some thoughts:
First:
Overall, I’ve updated that it’s too confusing to have any rate limits that restrict commenters’ ability to comment on their own posts. (There are 8 rules that create commenting rate limits, and previously only 2 of them rate limited you on your own posts). I personally think the 2 rules were reasonable, but I think they’ve mostly resulted in people forming inaccurate beliefs about how the rate limit system works. (i.e. they don’t realize it’s a rare exception for rate limiting to affect your own posts)
So, I’ve just shipped a code-update that makes it so there are now 0 ways to get rate limited on your own posts. (Well, one exception for the universal “you can’t comment more than ‘every 8 seconds’” rule, but that’s a pretty different rule)
Two:
I do basically think it’s correct for you to have been rate limited, for the reasons John/Shminux and others have described. You argued poorly about a topic that we’ve covered a lot, in a fairly inflammatory way. This is precisely what the rate limits are there to nudge people away from.
I do think that people get more downvoted on some kinds of tribally-loaded topics, and I wish that were different. I’ve tried to tune the rate limits so that they do correctly rate-limit-people when they make bad arguments and get downvoted, but don’t excessively rate limit people when they get “excessively tribally downvoted”. I’m not sure if I’ve exactly succeeded at this, but I’ve been tracking the rate limits that get applied over the past month and I think mostly endorsed how things shook out.
Three:
More positively: fwiw, I have felt like your other recent posts seem like your doing a fairly reasonable thought process, given your current epistemic state / skills. (This is based on a cursory glance rather than reading them in detail, I don’t know that I think all your mental motions make sense, but, like, it looks like you’re trying to think through and argue about various object-level issues in a way that seems healthy)
It’s perhaps worth noting that, whatever internal function people use to decide how to vote, the effect of their votes is to make up-voted things more visible and make down-voted things less visible on the margin, and this sends a feedback signal to incentivize writing things that get up-voted to the extent you want your writing to be seen by more people.
Under this regime, rate limiting seems fair to me, as it’s an extension of controlling how much marginal content you produce can be seen. I’m not sure if it’s perfectly well tuned, but seems likely that it’s having a useful effect of causing there to be fewer things on Less Wrong that Less Wrong readers don’t want to read, and even as salty as I get about it when I think my own insights are under-appreciated or misunderstood, I still accept that’s just part of how Less Wrong has to work given that, at least for now, it’s got to determine quality based on the aggregation of low-bandwidth signals.
I got rate-limited a few weeks ago for a small number of strong downvotes on a single comment. I blame the over-indexing on strong-votes, and still overall support the system. It DOES have some false-positives, but there is a real problem with otherwise-valuable posters getting caught in a high-volume useless back-and-forth, making the entire post hard to think about.
Rate throttling is a transparent, minimally-harmful, time-limited mechanism to limit that harm. It makes mistakes, and it’s annoying when one disagrees with votes. But I don’t know of a better option.
FYI we’ve since updated the system to only trigger based if there are enough unique downvoters on ‘net-negative comments’, which I think should reduce the false positive rate.
(Ie I think the reason it triggered in your case was that you also have some random downvotes on other upvoted comments)
Yeah, part of it was the selection for recency of vote, even on old comments—a positive-total comment from the past got some new downvotes, and that triggered the throttle.
That’s probably a flaw that shouldn’t result in rate-limiting (which reduces NEW posts, not old ones, obviously), but my main point is that the imperfect implementation is still pretty good.
I’m glad you’re continuing to refine it, but I don’t want it removed entirely or reworked from the ground up.
Lesswrong will probably have to become invite-only at some point (e.g. if Yudkowsky wins a nobel prize in 2035 or whatever) so the rate limit is a good middle ground for now.
For one thing, there is a difference between disagreement and “overall quality” (good faith, well reasoned, etc), and this division already exists in comments. So maybe it is a good idea to have this feature for posts as well, and only have disciplinary actions taken against posts that meet some low/negative threshold for “overall quality”.
Further, having multiple tiers of moderation/community-regulatory action in response to “overall quality” (encompassing both things like karma and explicit moderator action) seem good to me, and this comment limitation you describe seems like just another tier in such a system, one that is above “just ban them”, but below “just let them catch the lower karma from other users downvoting them”.
It’s possible that, lacking the existence of the tier you are currently on, the next best tier you’d be rounded-off to would be getting banned. (I haven’t read your stuff, and so I’m not suggesting either way that this should or should not be done in your case).
If you were downvoted for good faith disagreement, and are now limited/penalized, then yeah that’s probably bad and maybe a split voting system as mentioned would help. But its possible you were primarily downvoted for the “overall quality” aspect.
I share your frustration. I made ONE downvoted comment and then was restricted for quite some time.
I wanted to muse on this more:
Part of the back of my mind wondered if the comment was “bad enough” to be worth it, but then I realized it doesn’t really matter, because the tradeoff is average positive karma versus single negative karma. If you have an average karma of +3, then a single −50 will take an extremely long time to recover from, even if it was something silly like your comment being misinterpreted as hostile. On balance, I agree with you, but I thought it was worth sharing why.
Here goes my one comment for a day. Or may be not? Who knows? It is not like i can look up my restrictions or their history in my account page. I will have to make two comments to figure out if there a some changes.
One comment per day is heavily discuraging to participation.
If you’re going to come into an echo chamber of doom and then complain about being censored… well, what did you think they were going to do? It’s like walking into a Mormon ward and arguing with the bishops that Joseph Smith was a fraud.
The true believers are not going to simply convert because you disagree with them. The confirmation bias won’t allow that when they’re in a feedback loop.
They will gently instruct you to be more “intelligent” in your discourse. Of course, if it turns out they’re the “morons” then there will be a moment of amusing reflection when they’re still alive twenty years from now and AIs didn’t kill everyone.
”Damn it! Our fearless leader promised us we’d all be dead by now.” ;-)
If and when they don’t die by AI apocalypse they will then have to adhere to a new religion. Maybe aliens coming to take us away? At least that isn’t easily falsifiable by the passage of a couple decades.
Before everyone takes offense and begins writing their Senator, I don’t know if they’re actually morons, but they love to point out that those with whom they disagree must not be intelligent. Rather than entertaining the possibility that they’re the idiot in the room. At least as it relates to their existential risk of AI propaganda.
Their ad hominem attacks are shrouded with all of the window dressings of a religious zealot, “You might not get so many down votes if you stopped saying you disagree with our religious leader and instead reworded it to be so vague that we have no idea what you’re trying to say so that we can all just get long. When you say our leader is full of $%^* it makes us sad and we’re forced to shun you. “
I’m translating so what we’re all on the same page. =-)
I enjoy some of their rhetoric, in the same way I enjoy sci-fi stories. However, a dilettante shouldn’t fall in love with their own creative story telling.
This seems to be written from the perspective that life in 2043 will be going on, not too different to the way it was in 2023. And yet aren’t your own preferred models of reality (1) superintelligence is imminent, but it’s OK because it will be super-empathic (2) we’re living near the end of a simulation? Neither of these seems very compatible with “life goes on as normal”.
We don’t know for certain if all AI superintelligence will be empathetic (not all humans are empathetic), but we do know that it’s training on human data where that is an aspect of what it would learn along with all the other topics covered in the corpus of human knowledge. The notion that it will immediately be malevolent to match up with a sci-fi fantasy for no good reason seems like a fictional monster rather than a superintelligence.
It would have to be an irrational AI to follow the Doomer script who are themselves irrational when they ignore the mitigating factors against AI apocalypse or do a lot of hand waving.
It’s a scale. You’re intelligent and you could declare war on chimpanzees but you mostly ignore them. You share 98.8% of your DNA with chimpanzees and yet you to my knowledge you never write about them or go and visit them. They have almost no relevance to your life.
The gap between an AI superintelligence and humans will likely be larger than the gap between humans and chimpanzees. The idea that they will follow the AI doomer script seems like a very, very low probability. And if anyone truly believed this they would be an AI nihilist and it would be mostly a waste of time worrying since by their own admission there is nothing we could do to prevent their own doom.
We don’t know if this is a simulation. If consciousness is computable then we know we could create such a simulation without understanding how base reality works. However, a separate question is whether a binary program language is capable of simulating anything absent consciousness. The numbers cannot do anything on their own—since they’re an abstraction. A library isn’t conscious. It requires a conscious observer to meaning anything. Is it possible for language (of any kind) to simulate anything without a conscious mind encoding the meanings? I am starting to think the answer is “no”.
This is all speculation and until we a better understanding of consciousness and energy we probably won’t have a suitable answer. We know that the movement of electricity through neurons and transistor can give rise to claims of phenomenal consciousness but whether that’s an emergent property or something more fundamental is an open question.
That is not what most AI doomers are worried about. They are worried that AI will simply steamroll over us, as it pursues its own purposes. So the problem there is indifference, not malevolence.
That is the basic worry associated with “unaligned AI”.
If one supposes an attempt to “align” the AI, by making it an ideal moral agent, or by instilling benevolence, or whatever one’s favorite proposal is—then further problems arise: can you identify the right values for an AI to possess? can you codify them accurately? can you get the AI to interpret them correctly, and to adhere to them?
Mistakes in those areas, amplified by irresistible superintelligence, can also end badly.
I don’t think this is the case. For awhile, the post with the highest karma was Paul Christiano explaining all the reasons he thinks Yudkowsky is wrong.