Comment Permalink

DirectedEvolution2y171

My first thought is that I like these ideas. On its face, a rule-based temporary rate limit seems better than a mod-enforced permanent one.

Issues with karma- and post-count-based rate limits

Users can write zero-karma filler posts to avoid or end a rate limit that's based on the net karma of the last N posts.
It gives a lot of extra power to fast-acting activist or high-karma minority to squelch pushback with downvotes. If I know that arguing an unpopular view sets me up for downvotes, and that as a result, I not only will see my post collapse but may be seriously impacted in my ability to use the site, that makes me a lot less eager to make such arguments.

Even if you think LW doesn't have a problem with this now, I think putting this extra power in the hands of the silent downvote squad will change their behavior, and that it will change poster's behavior in anticipation of this risk. If this rule was in place, I could go through the profiles of LessWrong users, look for anyone whose last 5 comments had less than 8 karma, and strong-downvote them to eliminate their ability to use the site. When they posted their next post, I could once again strong-downvote it to block their ability to return for as long as I cared to do so.
There is an asymmetric justice issue here. The best thing that can happen if you get a lot of upvotes is you have a little more insulation against downvotes (except in the LessWrong review, where you might get paid on the discretion of the mods, which is correlated with upvotes). The worst thing that can happen if you get a lot of downvotes is you get rate-limited or banned, which is much worse. I'm not exactly sure how to deal with this - I doubt the money exists to just "pay people to write upvoted content," but of course I could be wrong.
Of lower import, we already suppress visibility of downvoted posts, and I think the visibility and status/emotional cost impact of downvotes is roughly proportional to the quality signal contained in the action of downvoting a post. Attaching a rate limit as well seems excessive.

Rate limiting people's ability to award upvotes and downvotes

I would feel much better about rate-limiting people's ability to assign karma to posts, restricting that power to more established (high-karma) users, and giving users more control over the visibility they assign to specific users, users with certain karma scores, and so on. I am open to severely restricting the ability to assign karma, much more so than the ability to write comments, and only then basing more severe punishments such as rate limiting on the basis of karma.

In general, if you're attaching more power to karma, then you have to more carefully regulate assignation of karma. If you're attaching more power to post frequency, then you have to regulate post frequency more carefully. Any such moves effectively empower the entire cohort of active and high-karma users with a fraction of moderator powers, and that marks a fundamental change in how LW is moderated. I think that needs to be done with much more caution than the mods simply giving themselves a new tool in the toolbox.

Rate limiting demon threads

Posts are in a tree structure, with the top-level post as the root, top-level comments as the first level of branches, and so on. It seems possible to calculate the net karma of any particular branch in the comment tree based on some rule like "median karma in this 2+ user thread is zero or negative" and rate-limit participation by all users just in that specific branch (or perhaps in the overall post). They might just jump to a different branch to continue their conversation, but I tentatively think it might help. You could potentially impose broader rate limits on users who consistently are making negative-karma posts in multiple threads tagged in this way as "algorithmically determined demon threads" to prevent people from spawning endless new demon threads when the old one gets rate limited.

I think that with this additional layer of restriction, we have limits of time, "conversational space," and level of restriction all softening the impact of rate limits. This is where I start to get more comfortable with linking the imposition of rate limits to karma and post count.

See in context

48 "Rate limiting" as a mod tool

by Raemon

23rd Apr 2023

5 min read

48

A lot of my recent thoughts on moderation have included an emphasis on "rate limiting as a primary mod tool." This is very much in the "Ray has a hypothesis and it might not work out" phase, but I think it has promise.

The general principle is "moderators should be able to employ the least-amount-of-constraint/force possible to enforce a decision or give that decision teeth." A full ban is a pretty extreme option, and being limited to that option makes me much more hesitant to take any kind of action against a person.

A rate limit is a pretty flexible tool. You can rate limit comments, posts, or (possibly) wordcount. You can limit by hour, day, week or more. It can be limited to a specific post, or across posts in general (i.e. N comments per post per unit-of-time), or sitewide (N comments total per unit of time)

But at least some people have reported experiencing limits as very costly, and have described it as "being banned." Ruby has argued it might basically just feel like moderators are trying to ban people without really acknowledging the magnitude of what they're doing.

I'm curious how people feel about it. I want to outline a bit how I'm thinking about it.

There's roughly 3 ways I think I want to use rate-limits

1. As a soft, temporary tool to give moderator-warnings a little more teeth.

If someone(s) is/are spiralling into a demon thread, we could give them a temporary "1 comment per ~hour" rate limit on that particular post, to force them to slow down and chill out. If they've done something that seems particularly norm-violating, we can give them a temporary significant rate limit. The alternative here would be to do nothing, or give them a temporary 1-week ban. It seems at least in some cases better to do the rate-limit, although I could imagine in some cases "taking an actual break for a week" would be better.

2. As a response to serious problem users who are either new, or established but have been given previous warnings.

The alternatives are "permaban," "doing nothing", and "moderators consistently have to be-on-top-of-things giving 'temp-bans' or other local-reactions to serious problem users each time they do something bad."

Doing nothing means leaving the garden untended, falling into entropy. "Moderators have to keep paying attention" is just... way too much work. From this perspective, a significant rate-limit is harsh, but the alternative seems probably worse. Rate limits give the user an option for proving they can change their commenting habits.

A thing that might change my mind here is if in practice, either most users just prefer exile to second-class-citizenship, or if longterm rate limits just in practice turn out to be a trap for people who really should move on.

3. As a default-state for new users.

New users sometimes show up and start commenting prolifically, before they really understand the site norms or background knowledge.

I think it's a pretty reasonable position for "Users start out with one comment per day. If they get even slightly upvoted, they get a bit more commenting permissions, and if they get, say, 5 or 10 karma they become unfettered." I predict that, once the site fully shifted over to this, it'd feel fairly reasonable and gamified in a positive way.

There's an awkwardness right now of "there's a lot of users I would have wanted to start out more rate-limited in this way, but who showed up in the past year or 2, who've been making lots of downvoted comments and not really learning from it." So there is an awkward/painful readjustment period for those users.

The main alternative to this (which is what we're actually doing right now, as of a couple weeks ago), is manually approving each individual comment from new users, and mostly not letting them post unfettered until they've posted at least one reasonable-looking one.

The other alternative is, of course, to just let people start posting/commenting willy nilly, but I think that's mostly bad for both existing users (who get a much lower signal/noise ratio) and new users (who get downvoted and criticized a lot).

(note: I previously wrote about this here)

Potential risks/costs/problems

Problem 1: You can't see which people are rate limited, and that makes conversations confusing.

Potential solution: Add a small icon on comments that conveys "this user is under a rate limit and can't respond right away". (potential problem with solution: feels too much like a Scarlet Letter? But I think it can be designed in a fairly subdued way that seems mostly utilitarian.

Problem 2: Maybe even with the icon, it's just a lot of cognitive overhead for users keeping track of how rate limiting affects conversations?

I don't know how big a deal this would be, and presumably it depends on how often it's used. If it turned out to be a big deal in practice, it might be a major update for me.

Problem 3: Some authors might actively like users that are under a rate limit.

Potential solution: give authors a whitelist tool that lets them allow particular friends/colleagues to comment unfettered on their posts (or, on specific posts)

Problem 4: The new-user-setting might be too strict. Vladimir Nesov was worried about this, saying "There should be fast tracks that present no practical limits to the new users. First few comments should be available immediately upon registration, possibly regenerating quickly. This should only degrade if there is downvoting or no upvoting, and the limits should go away completely according to an algorithm that passes backtesting on first comments made by users in good standing who started commenting within the last 3-4 years."

We have done some backtesting like this (looking at the early comments from good users), so we have a rough idea of what to expect here. There were some good-users-by-post-people's-lights who got downvoted in their first few comments.

I'm... not sure I buy this is that big a problem? Obviously if this caused good users to give up on the site and leave, I'd consider that a major loss. But that's not my first guess of what would happen – I think potentially-good-users would mostly see "oh, this place actually has standards. Lemme try and meet the standards and get in", and maybe that takes them an extra day or so? That seems fine to me.

But if I were persuaded this was more prone-to-people-giving-up I'd update a lot.

(Currently we're doing manual approval of new user comments which is I think mostly an even stricter regime than this implies. This user found it explicitly good to get rejected rather than subjected to downvotes and annoyed criticism. I'm not sure how common an experience that'll be, and am unsure how to integrate this all into a new cohesive vision for new-user moderation)

...

Anyways, those are some thoughts about rate limiting. I am curious how people feel about this compared to other options.

Moderation (topic)Site Meta

Personal Blog

48

Mentioned in

83Automatic Rate Limiting on LessWrong

New Comment

36 comments, sorted by

top scoring

Click to highlight new comments since: Today at 11:56 PM

[-]ChristianKl2y3025

Historically, a lot of how the garden of LessWrong was tended was through karma votes. I think a system that builds on karma votes is better than one that requires a lot of manual moderator calls.

How about a rule like: "If the median karma without counting self-upvotes of your last 5 messages is net negative you are rate limited to one message per 6 hours", "If the median karma without counting self-upvotes of your last 10 messages is net negative you are rate limited to one message per 12 hours" and "If the median karma without counting self-upvotes of your last 20 messages is net negative you are rate limited to one message per 24 hours" could work for existing users as well.

This means that even established users who write three net negative comments in a demon thread get rate limited to 6 hours, but I think that's fine and if the rules are clear I wouldn't expect an established user to leave because of it.

For new users that would mean that by default they don't have a rate limit but if anyone downvotes their first message they automatically have the 24-hour rate limit and the steps to getting rid of the rate limit are obvious.

[-]DirectedEvolution2y171

My first thought is that I like these ideas. On its face, a rule-based temporary rate limit seems better than a mod-enforced permanent one.

Issues with karma- and post-count-based rate limits

Users can write zero-karma filler posts to avoid or end a rate limit that's based on the net karma of the last N posts.
It gives a lot of extra power to fast-acting activist or high-karma minority to squelch pushback with downvotes. If I know that arguing an unpopular view sets me up for downvotes, and that as a result, I not only will see my post collapse but may be seriously impacted in my ability to use the site, that makes me a lot less eager to make such arguments.

Even if you think LW doesn't have a problem with this now, I think putting this extra power in the hands of the silent downvote squad will change their behavior, and that it will change poster's behavior in anticipation of this risk. If this rule was in place, I could go through the profiles of LessWrong users, look for anyone whose last 5 comments had less than 8 karma, and strong-downvote them to eliminate their ability to use the site. When they posted their next post, I could once again strong-downvote it to block their ability to return for as long as I cared to do so.
There is an asymmetric justice issue here. The best thing that can happen if you get a lot of upvotes is you have a little more insulation against downvotes (except in the LessWrong review, where you might get paid on the discretion of the mods, which is correlated with upvotes). The worst thing that can happen if you get a lot of downvotes is you get rate-limited or banned, which is much worse. I'm not exactly sure how to deal with this - I doubt the money exists to just "pay people to write upvoted content," but of course I could be wrong.
Of lower import, we already suppress visibility of downvoted posts, and I think the visibility and status/emotional cost impact of downvotes is roughly proportional to the quality signal contained in the action of downvoting a post. Attaching a rate limit as well seems excessive.

Rate limiting people's ability to award upvotes and downvotes

Rate limiting demon threads

[-]Raemon2y50

Users can write zero-karma filler posts to avoid or end a rate limit that's based on the net karma of the last N posts.

Note: they can't do this if ending the rate limit requires positive karma, which is why the threshold is more like +.5 karma per comment than 0.

[-]ChristianKl2y40

I think looking at the median is generally better than the mean. I don't think a single highly downvoted message should trigger rate limits even if it turns the mean under 0.

[-]Raemon2y20

I'm not sure median is great either, but I agree with the point of "one highly downvoted comment shouldn't trigger it, and shouldn't be that hard to dig yourself out of".

I think the right function probably looks at number of net-downvoted comments, neutral comments, and comments that gained 2+ karma or something, and triggers based on the ratio.

[-]ChristianKl2y20

I used the term message above as hypernym for post/question/comment/shortform. Do you think that it should focus on comments or are you also thinking that all types of messages should count.

Maybe the function should be "median of last X messages that don't have neutral karma"?

[-]Raemon2y40

I meant to be using 'comment' also as a more generic term, sorry for confusion. (I do think there are important differences between comments and posts, though, so I wouldn't necessarily treat them interchangeably here)

[-]DirectedEvolution2y20

That's true. My main concern is really the amount of power this places in ordinary users' hands to suppress the activity of other users, and your point here emphasizes just how much power that really is.

[-]ChristianKl2y20

I could go through the profiles of LessWrong users, look for anyone whose last 5 comments had less than 8 karma, and strong-downvote them to eliminate their ability to use the site.

I would expect that this would quite often result in other people upvoting those downvoted posts. But you are likely right that you don't want the votes of a single user to be able to start the rate limits for any established user.

[-]gilch2y20

I would feel much better about rate-limiting people's ability to assign karma to posts, restricting that power to more established (high-karma) users, and giving users more control over the visibility they assign to specific users, users with certain karma scores, and so on. I am open to severely restricting the ability to assign karma, much more so than the ability to write comments, and only then basing more severe punishments such as rate limiting on the basis of karma.

I strongly don't agree with this one. If anything, I think people should vote (honestly) more often. It's a valuable signal. If it weren't this easy, we wouldn't get the feedback at all. If I got rate-limited on voting, I'd still be reading the same number of posts, but it would be too much extra work to budget my allocation to the most important things I'd read, so I wouldn't bother.

I do worry about karma getting gamed. We've already had an issue with that, with one user mass-downvoting unrelated old posts of another as a means of attack for a current dispute. The mods are on top of it for now, but I could imagine other exploits we haven't thought of popping up. I just don't think rate limits on votes are the answer.

[-]DirectedEvolution2y42

Hm. In theory, I agree with you - I think a good equilibrium (ignoring the cost of taking the time to vote on stuff all the time) would be lots and lots of voting to express views on how worthwhile the content is.

Let me articulate a little bit more of how I perceive karma being used in practice on LW.

Sometimes, it's a scout-mindset way of saying "this is a worthwhile comment. It should be more visible, I think people widely agree on that, and I'm going to help it out by upvoting." It's a similar function to what Kriss describes as the old function of hipsters. This is good and I would want this sort of voting to continue unrestricted.
Sometimes, it's a soldier-mindset way of saying "I'm on the side that's for/against this point, and I have to up/downvote it lest it appear that the other side is more popular!" In soldier-mindset situations where the post supports the majority view, this looks like a lot of upvotes, but with many more total votes than upvotes. If the post supports a minority view, this looks like a post with a decent number of downvotes, a couple upvotes, never getting much oxygen because the soldier-mindset majority has inevitably suppressed it. Overall, this seems bad for site epistemics, and I would want to see zero or rate-limited voting in these situations.
Sometimes, it's a more intimate way of expressing warmth/appreciation/encouragement or saying "I read this," or coldness/discouragement, where you often can guess exactly who upvoted/downvoted you. Here, upvotes seem good (a signal of a valuable discussion), whereas downvotes seem bad (by creating bad feelings while not actually terminating an unproductive discussion, and potentially catalyzing demon thread formation).

I think that if LW implemented rate-limited voting, we'd still get the scout-mindset form of voting, and my intuition is that most people, especially established users, would mainly choose to spend their upvotes on rewarding quality content, punishing unusually bad content, and encouraging conversations they'd like to have. And they would reallocate votes away from sides-taking contests and demon threads.

I could be very wrong about that, but it's these intuitions that make me support rate-limited voting.

[-]M. Y. Zuo2y10

It gives a lot of extra power to fast-acting activist or high-karma minority to squelch pushback with downvotes. If I know that arguing an unpopular view sets me up for downvotes, and that as a result, I not only will see my post collapse but may be seriously impacted in my ability to use the site, that makes me a lot less eager to make such arguments.

This is a solid point.

I've found many interesting comments and posts from many folks on LW that weren't too far from neutral, i.e. -10 < karma < 10 (after excluding low effort trolling and such)

Where the highly upvoted stuff is usually far more predictable and not nearly as interesting.

The exact ratio depends on how popular the topic is overall, how long it stayed on the front page, etc., but I would agree that punishing the bottom half would almost certainly reduce the number of these intriguing writings.

It might also cut down on the low-effort posts so I'm not certain it would be a net negative. Maybe this requires a gut-feel decision by the moderating team?

[-][anonymous]2y*103

Note you have essentially banned me in that I would be treated better as a new user. Your forever 1 per day limit - you obviously have no jira or system to review my account, nor did you do any numerical analysis (UPDATE: you provided https://www.greaterwrong.com/users/gerald-monroe?sort=top&show=comments page, which I believe highly supports my points) to find the many highly upvoted posts I made in the past - puts me in a situation where as a new user I can just farm a little karma and have an unrestricted account.

This makes the punishment from my perspective the same as a ban. Either way, my recourse is to create a pseudonym/use a proxy, or endure your (in my opinion) unjust punishment.

Obviously I have not done so or you would not be reading this post - I have taken the opportunity to focus on work - just noting this is the outcome of your policy decision.

I will admit that the last few posts I made might not have been up to your standards - the discussion over AI pauses, which have enormous real world costs, got heated and I tanked a lot of downvotes. Obviously in my opinion my standard of argument was appropriate to the level of discourse of the discussion - Eliezer himself blows off responding to criticism, numerous users claim a superintelligence can essentially ignore the laws of physics including material and thermodynamics and compute substrate limitations and take over the planet within hours. Hardly a rigorous debate.

UPDATE:

Since I'm not allowed to respond to habryka's post, I am saying that I have already been punished with a de facto permanent ban. So threatening me with more ban if I were to create a new account in an anonymous way, is not a useful threat. Obviously a second permanent ban can just be evaded by... and so on. I have not done this, but other users no doubt will respond in this way.

My other comment: Ben's search https://www.greaterwrong.com/users/gerald-monroe?sort=top&show=comments show's 99 comments over 5 karma. That's not nothing. Or 10% of every comment I made. There are 19 pages at 20 comments a page with positive karma, or 380 comments that at least 1 net user upvoted.

Maybe I numerically deserve this punishment but a new user with even 1% of what I posted would probably be given unrestricted access.

[-]habryka2y76

Just to be clear, creating alt-accounts to avoid rate-limiting or bans are I think sufficient reason for a permanent ban. Please don't do that.

[-]Raemon2y30

First I do just acknowledge that the rollout of rate limits hasn't been optimized for being as good-an-experience-as-possible for users (largely due to time constraints, and the quality of site-comments feeling like an overall pretty urgent thing last week. A lot of our current choices has been due to a general wave of poor quality AI comments.

In particular, I think it was bad of us to not give you more of a sense of how/when the rate limit might be lifted. I actually had started a reply to your last PM explaining our method a bit more, and then got distracted and didn't send it, and I'm sorry about that.

I did expect, at the time, to fairly quickly ship an update that makes all new users start with a rate limit of 1/day. We haven't ended up shipping that yet for a variety of reasons, but I still expect to in the not-too-distant future. (which would mean you wouldn't be under harsher restrictions than a new user)

The rough algorithm we've been following is:if we notice a newish user we don't recognize has an average karma-per-comment of <1.5 in their past 20 comments, look into their individual comments. If they seem not meeting our quality bar, give the user a 1-per-day rate limit. Our quality bar is higher for AI content right now because there is such a deluge of it.

(We picked the "1.5" flagging threshold based on looking at a bunch of comments from various users, seeing checking the average, and seeing where our bar of "okay we basically feel good about this user being net positive" tended to lie)

My idealized version of this includes automatically removing the rate limit if the user makes enough upvoted comments/posts, and also gives some advice on how to improve. We're working on that, and I'm sorry about the current situation.

We give a lot more leeway to users who've established themselves with solid posts and comments. I'm expecting a typical person meeting the "extra leeway tenure bar" as something like 3+ posts with 75+ karma (on topics that aren't optimized for karma-farming, such as certain kinds of drama posts). I'm not quite sure what the number of Good Comments would be to meet the bar, but don't think you're above the bar yet.

At the time I write this, your average karma for your past 20 comments, including your self-upvotes, is .9 (and was lower at the time we set the rate limit. Since this includes your self-upvotes, it means your recent comments have been net-downvoted. I'm not saying average karma is a perfect measure of quality here, but I do think it correlates reasonably well. My impression when I looked over your comments at the time was that many of them seem to be making more assertions than arguments and being kinda lowkey aggro about it.

[-]the gears to ascension2y3-4

I would request @Gerald Monroe have at most 2 hour rate limit; their comments, while sometimes frustrating due to disagreement, represent what I see as an important and underrepresented view: the "lol I know how to build it and it's really obvious" view. Currently I only know a few others to have a similar perspective. I want to be able to discuss with representatives of that view; censoring them all would be frustrating. Certainly not all of them actually know exactly, but I don't think any of the ones I'm thinking of are entirely wrong in their hubris.

[-]Raemon2y30

I think my preference here is that that kind of conversation is allowed but doesn't come up in every object-level post. (I'm not sure how to actually operationalize this as a moderation call)

[-]Thoth Hermes2y10

I suppose this location would be as good as any for my response to this issue (I note that I currently have a 1-post / 3-day limit on my account, so this means this comment should be fairly long).

Looking back at my own top comments, I have noticed that what LessWrong the community tends to favor are things I would consider to be more critical in general: Not things that "disagree" on equal-footing so much as critiquing a post coming from the position as more knowledgeable and more experienced than the person I am critiquing. This means that if I wanted to "farm karma" so to speak, I could do so by limiting my comments to criticisms that are flavored with a wee bit of condescension.

When looking at what others' consider to be your best contributions, it presents somewhat of an awkward question: Are these the same things I would rate as my best work? If not, does that mean that I was trying less hard, or taking it less seriously than I should when I wrote them? We have to consider that when one receives feedback on their work, positive or negative, one must interpret it by their own lights in order to incorporate that feedback in a way that would be considered adequate (assuming that they do indeed wish to improve themselves with respect to overall karma).

Note that it would be sort of funny if you sent a message to someone (in moderator capacity) in something like the following way:

"[User], you've been posting fairly frequently, and my subjective impression as well as voting patterns suggest most people aren't finding your comments sufficiently helpful. For now I've given you a 1-per-day rate limit, and may evaluate as we think more about the new standards. As far as feedback goes, do you think you could try and be a little more critical to people, and flavor your posts with a wee bit of condescension? That would really help make LessWrong the well-kept garden it aims to be!"

In other words, optimize not for being "low-key aggro", but rather for criticisms flavored with a wee bit of condescension. You don't want to sound like the Omega, who can't rule out that he truly deserves to be at the bottom of the status-hierarchy, just because he received some negative feedback and perhaps some moderation warnings. It's easier to perform "established group-member" by limiting your output to critique that aims to come across as being helpful to the community as a whole, rather than the person you're delivering it to.

The only problem with that is that when I am trying to write a post and I want feedback that's not just from my own judgement, my mind turns to "what would people upvote more" and my own experience thus far tells me that if I were simply to alter the tone of my posts / comments as opposed to the actual content of the posts / comments, I could alter the resulting approval rating substantially more than I would otherwise think it deserves.

It's not that I disagree with or dislike my own comments that I believe cater to the community's expected response so much, just that I think that optimizing for that would be necessary to avoid the restrictions placed on my commenting, and I think optimizing for it enough would bring me outside of what I would consider to be my own standards and judgements for what I actually think as well as how expressing it should actually be done.

As a moderator, you have the responsibility of tuning the dials and knobs that change whatever metrics the users are optimizing for, and in this case, the weight applied to 'impressing just the community' as opposed to 'just speak your mind, using your own standards for what you believe qualifies as good for yourself as well as the community' (or equivalently, their relative ratio). You have to be pretty sure that increasing that weight is what you want. That weight applies to everyone, of course. So if it is tuned too high, then you get a situation in which everyone is optimizing for what they think everyone else thinks is good.

For the record, I don't think that weight should be zero, but I also think that it will be non-zero somewhat naturally, so that any weight increases you apply to it with the infrastructure, rules and norms might look like they are being added to zero, but are actually being added to a non-zero initial quantity. It may be that you come to the conclusion that the optimal amount is higher than whatever you deem to be the initial amount, and so that some restrictions are still good to have. Please consider possibly subtracting out the base value from your estimations of how much stricter you want the norms on the site to be.

[-]Viliam2y42

I checked your comment history. The top comments at this moment start with:

Here’s why I don’t find your argument compelling (K 15)
These norms / rules make me slightly worried that (K 13)
Here’s why I disagree with the core claims of this post (K 7)

Sounds like evidence in favor of "disagree with people, get upvoted".

On the other hand, your comments with karma below zero:

I’ve never enjoyed, or agreed with, arguments of the form (K -1)
I think that ... would imply that ... Personally, I think it’s pretty easy to show that ... is wrong. (K -1)
I don’t think that ... means that ... (K -1)

So now it seems more like you disagree a lot (nothing wrong with that), and some of those comments get upvoted, and some of them don't. The upvoted ones do not feel more condescending than the downvoted ones.

[-]Thoth Hermes2y10

Actually, my lowest three comments are:

It seems to be historically the case that “doomers” or “near-doomers” [...] (K -9)
AFAIK, the Secretary-General is a full-time position, e.g., [...] (K -5)
Remove the word “AI” and I think this claim is not really changed in any way. AI systems are the most general systems. [...] (K -5)

The following is simply my own assessment of why these comments were downvoted. For the first one, I assume that it was because of the use of the term "doomers" in a pejorative sense. (This is closer, I believe, to what I called "low-key aggro" in my earlier comment.)

I am not sure why the second one was taken so poorly, and I imagine that whoever downvoted it would probably claim it to be snarkier or more disrespectful somehow than it actually was. This is unfortunate, because I think this serves as evidence that comments will often be downvoted because they could be interpreted to be more hostile or low-effort than they actually are. Alternatively, it was downvoted because it was "political."

The third one is also unfortunate. Disagree-downvoting for that comment makes sense, but not karma-downvoting. If you were to counter that it was somehow 101-material or misunderstanding basic points, I would still have to strongly disagree with that.

My second-highest comment is about why I am worried about site-norms unfairly disfavoring discussions that disagree with major points that are commonly accepted on LessWrong or taken as catechism, so that should also support the idea that if such norms exist, you will observe that comments that do so also appear to be karma-downvoted, so as to limit their visibility and discourage discussion of those topics.

This still supports my main point, I believe.

[-]habryka2y20

nor did you do any numerical analysis to find the many highly upvoted posts I made in the past

Just to check, you have 4 posts on your profile, none of which have above 10 karma. Are there any posts I am missing?

[-]Ben Pace2y20

If you search by top-comments (as you can on GreaterWrong) you can find this 55 karma comment by Gerald Monroe from 2 years ago.

(The next highest karma comment is 22, 20, 19, so there's only the one comment I would call 'highly upvoted'.)

[-]habryka2y20

Sure, though I was specifically asking about posts, which Gerald mentioned.

[-]Ben Pace2y41

I think Gerald is using 'posts' to mean any sort of content that has been 'posted', like when he writes "Obviously I have not done so or you would not be reading this post" referring to the comment he has written.

[-]habryka2y21

Ah, that makes sense.

[-]Eli Tyre2y*93

I haven't read the whole post yet.

My initial thought: I'm much more sketched out by platforms rate limiting users, than I am with them straight up banning them. The first is much more "micro-managing-y", and I wonder if it can lead to more subtle and powerful distortions than outright bans, which are at least transparent.

I haven't thought about it much, but I think I feel much safer about a marketplace of ideas that has a gate (some people are not permitted) than a marketplace of ideas that is slanted to advantage some ideas over others, especially if group deciding the slanting is centralized.

I think part of my intuition here is that, if you don't let people into the marketplace, that's one thing, but if you fuck with which things get propagated, you are distorting the basic mechanism of the marketplace of ideas, in which people sharing the things that they think are true leads to correct ideas gaining dominance over time.

On the other hand, given social media and hyper-competive memes, I'm pretty sympathetic to the idea that "the marketplace of ideas" is an outdated frame, and we need better concepts for pursuing the enlightenment project in a way that doesn't get subverted.

[-]Screwtape2y63

A second order effect I'm a little worried about is rate limits create an incentive to mix threads.

If there are two people who make different points in comments and I want to respond to them, ordinarily I would reply to each of their comments and the discussion stays organized. If I'm rate limited and there are two people I want to respond to, maybe I write both replies in a single comment under one of them and mention the other.

One answer to that is asking people not to do that. One answer is eh, it's not that bad, sometimes that kind of mixing happens even without rate limits and it's fine. Another might be doing word count limits instead of comment limits.

I don't think this is a big issue. I did think it was worth mentioning while you were looking for feedback.

[-]DirectedEvolution2y60

List of thoughts:

I have no intrinsic problem with the tool of rate limits. It’s more a question of when it’s the best tool for the job and which jobs to prioritize.

Karma already does a lot of the job of deprioritizing uninteresting or poorly thought out comments. Its failure mode is that it’s a status modifier and can lead to hurt feelings, since somebody had to make the decision to click the downvote button on you. Likewise, a poor comment that is worded in a hurtful way can still lead to hurt feelings, no matter its visibility. I’d love to be able to turn off karma score visibility entirely, and just keep the visibility ranking they generate. I don’t like greaterwrong’s design but I do like how it hides karma scores.

A second issue is that you can moderate disruptive active users, but it’s hard to recover a user who quits in frustration. Older users have more connection to the site and more awareness of its norms and how to make it work for them. So I would prioritize using new tools to make the site more welcoming for newcomers, and educating old timers on the tools that exist to make the site work better. I’ve been on the site for years and didn’t know you could block specific users until a few days ago. Making all these options very visible and easy to use, all in one place, and expanding their number seems great. Periodic reminders would be a good idea.

Here’s a weird idea. I should be able to tag specific users with standardized “reputation badges” that nobody else can see directly - could be just a user specific upvote or downvote that only affects their visibility to me. Could also be something like “often rude, sometimes insightful.” I’m even more interested in a sort of LW opt-in whisper network: I could take user X with a “hurts my feelings often” standard tag, and then could see how many other users have tagged them that way, or something like that. Even minus the social element, giving me a way to make an assessment of a specific user that will pop up alongside their comments (but only visible to me) would be somehow helpful. Hard to explain why exactly.

[-]gilch2y53

I approve of the rate limiting idea, especially considering the influx of new users, for Concentration of Force/Eternal September reasons.

Given that framing, and to make it feel like less of a punishment, and to prevent easy evasion with new accounts, I think new users should be rate limited by default, and the mods should be willing to dynamically adjust these defaults depending on the number of new users. If it's just a few at a time, maybe we can assimilate them without rate limiting them at all. When it's more, impose the rate limits on everyone below a certain karma threshold. When we're getting overwhelmed, to prevent an Eternal September scenario, the allowed posting rates could be much lower than usual. Keep watch on the numbers.

More speculative, but I could imagine something like a green card lottery or sponsorship system to help us assimilate users at a more manageable rate, even when demand exceeds our capacity to do so, although increasing that capacity is also worth looking into. About the best I can do now is point at The Sequences.

So, for example, the mods could select a certain number of promising new users per month and remove their rate limits. Or that number could be selected at random from new users who apply for the privilege. Or with a low enough karma threshold, maybe new users would graduate to assimilated status at a reasonable rate on their own, but this would also have to be tuned by the mods based on the current numbers.

[-]Ruby2y50

Ray covered some of this in the post above, but share some of my own thoughts:

Why be interested in rate limits

I think we got on to the idea of rate limits because of the shortcomings of bans:

Bans feel like a big deal. This means that us moderators are reluctant to issue them, and when we do, we spend a lot of time on them (deciding if it's right, justifying it, etc.)
Bans don't give you a way to redeem yourself, particularly where the issue a user is putting out a large amount of really mediocre (but not exactly norm-violating) content.
You can give a temporary ban of pretty short duration, see how goes, reapply, etc....but then you've just re-invented rate limits.

If I recall correctly, the teams interest in rate limits is closely tied to an interest in "automoderation" based on various rules. The idea is that under certain conditions, e.g. a user's karma is negative, rate limits get applied. Rate limits have the nice property of being gradual and of letting your redeem yourself. They're less of a big deal, so less scary for mods to apply, and if you tie them to a well-tuned automoderation, they're a scalable way to keep quality high and restrict users who pull the site quality down with large amounts of meh to actively bad content.

(Karma on every individual post helps but I feel is insufficient and inefficient. 1) for karma to be the basis of what's visible, still someone has to look at the low quality content, 2) seeing low quality content even if you don't read it, still makes the site feel lower quality and like it doesn't have standards, 3) it takes design/engineering work to ensure low quality stuff doesn't end up "leaking through", 4) Getting downvoted doesn't necessarily come with any explanation of why you're getting downvoted. All these reasons make me interested in means of maintaining standards beyond just having people vote on each thing. For example, if I user gets negative karma on all their stuff, maybe we should have them produce less stuff.)

Possible problems with rate limits

I don't think we've used rate limits enough to be really be sure how real these are, but I think we need to think about them. Here are some:

I think many of these can be addressed with good design, but we'd have to get them right.

Possible costs on other users
- In comment threads, I could see it being frustrating to interact with someone who is rate limited. Perhaps they wrote something and you asked a clarifying question, but they can't answer for 24 hours. And maybe they won't because they won't prioritize responding to you over something else tomorrow. This is a problem even if you know they're rate limited until some time.
  - There'd be questions of do we just signal when someone has exceeded their limit, or even before they do so (as a sign they might not reply).
- It's just weird to have a whole class of people who can't act freely on the site. With a ban, a person is just absent. Here, you might be frequently reminded of the restricted "2nd class", "being punished" users.
- I'm guessing some number of rate-limited users will complain. Do moderators ban that? (Kinda dictatorial) or do other users have to see those complaints periodically.
- Confusion of who is facing what restrictions where, e.g. users who are generally rate limited but have exceptions in some places.
More confusing than a ban
- With a ban, you're just unable to post and comment. Simple. If you ban expires, you can get back to it. But a rate limit, especially for commenting, could require this unfun mental rationing of your limit. "Oh, I have a thing to say there, but maybe I'll read another post and want to use it there, so I won't say this thing."
  - Personally, this'd likely just mean I didn't even think about using what limit I have.
- There's a question of UI. Do we have some salient thing that says "you are rate limited until..." or "N comments remaining" or something else that feels very punitive or something. I wouldn't like seeing that, but also seems like UI that you need.
Misapplied and unmonitored could be bade
- A think I'll have the team do before we proceed is build something like a dashboard over user currently rate-limited, so we can track whether it seems reasonable or not.
- I think we need to tread cautiously (and we will) with any rate-limited that happens as a result of an automatically applied rule. False positives here could be pretty bad and have a large chilling effect.

The details matter

I think there are cases where rate limits will work great to encourage good behavior or discourage users who contribute low quality stuff away from the site without it costing moderators a lot. For example, rate limits on making posts for new users will likely work pretty well. As an alternative to bans, we'll have to get more of a feel for their effects.

[-]Vladimir_Nesov2y*42

I think potentially-good-users would mostly see "oh, this place actually has standards. Lemme try and meet the standards and get in", and maybe that takes them an extra day or so? That seems fine to me.

All reasonable negative selection is like that.

[-]gilch2y30

More random thoughts. I could see editing comments as a way to evade rate limits, as happened here. We don't particularly have a norm against it yet, so I'm not really complaining about this particular occurrence, but consider what would happen if this pattern were to become widespread.

Demon threads might become epistles with lots of quotations. They might become harder to follow, and we'd only be able to vote on them once as a whole. This kind of seems like the same problem, but with a worse UI, although it might at least limit the blast radius.

Limiting word counts seems even worse. I think it's fine to edit to correct typos, or to reword things to clarify the original intent. It's also maybe acceptable to make additions if they're done very quickly or even later if labeled as such (but I question even that for rate-limited users). But if the length of the post was limited and the number of comments were limited, I could see the original comment getting edit-replaced with the reply altogether. Do we really want it to work like that?

It seems dishonest to me to completely change a comment like that. Other sites show old edit revisions to mitigate this sort of abuse. If LessWrong has that feature, I haven't found it yet. But if we had it, in combination with the rate and length limits, then we'd have to dig through the edit history to read the conversation. Again, same problem, but with a worse UI.

I suppose rate limiting or time limiting edits is a possibility. I don't particularly like it. Five minutes or something (a common edit limit time on other sites) often isn't enough time to proofread a longer comment. I don't think there's a way to preview how the markdown will render short of posting it (and even if there were, I'd probably miss things until after posting). I sometimes spot typos in my comments months later and correct them. The ability to freely edit like this is one of the things I like about LessWrong. It seems like a simple rule against abusing edits, and ways to notice (history) and report it (flags) and sanctions for violating it (Bans? Harsher rate limits?) is the best solution here.

Given that, the next workaround is probably a link to some other forum. I suppose we could try to block this somehow, but moving the conversation off site seems fine.

[-]Adam Zerner2y30

Cool idea. I applaud the creativity. Some thoughts:

Thank you for making various cruxes explicit.
It's one thing to implement this as a permanent thing and never look back, but it's another to try it for a month as an experiment and re-evaluate from there. I assume the latter is what would actually be done in practice and I think the bar for the latter is much lower than the bar for the former. But I think it can be easy to slip into having a bar closer to the former when evaluating whether this is worth trying.
I'm wondering whether other communities have tried something like this before and if so what can be learned.
I see moderator time as very valuable and something to optimize for.
Something to be careful of: giving more options to moderators may lead to a paradox of choice type of situation where having an additional tool to consider, rate limiting, now always makes you have to wonder whether rate limiting would be a better choice than whatever the alternative would have been.
As AI becomes more powerful I expect more and more users to come to LessWrong. I think it makes sense to explore ideas and prepare for that now.

Other responses:

But at least some people have reported experiencing limits as very costly, and have described it as "being banned." Ruby has argued it might basically just feel like moderators are trying to ban people without really acknowledging the magnitude of what they're doing.

Personally I wouldn't see it that way and would be surprised if more than 20% of people do.

Add a small icon on comments that conveys "this user is under a rate limit and can't respond right away".

Seems fine to me.

Maybe even with the icon, it's just a lot of cognitive overhead for users keeping track of how rate limiting affects conversations?

I don't see why it would. 1) Extended back-and-forths between two people seem like more the exception than the rule, 2) it's common for the turnaround time to be 24+ hours anyway, 3) I wouldn't expect too much rate limiting to happen, and thus it wouldn't affect too many conversations.

Some authors might actively like users that are under a rate limit.

I don't see that happening very often. When it does, I expect it to be pretty minor and not worth spending time on. If someone that you like is rate limited it doesn't seem like too big a deal to wait a little bit for the rate limit to expire.

I think potentially-good-users would mostly see "oh, this place actually has standards. Lemme try and meet the standards and get in"

Well put. I agree.

[-]Elizabeth2y30

it's common for the turnaround time to be 24+ hours anyway

I don't think we should assume the wait for replies for rate-limited users will be exactly the refresh rate (as in, someone given 1 post every 24 hours will always respond within a day) if the limits are site-wide, because people might want to participate in more than one conversation at time. I don't necessarily think that's a problem, I think slower conversations are probably an improvement and it would be weird to say a rate limit is good except it reduces the number of comments people can leave, but I do think it's important to track that the speed reduction may be greater than planned.

[-]the gears to ascension2y20

sounds great! What's the rate limit on mods using it? ;)

[-]Shmi2y20

My guess would be that starting lightly and locally and iterating would be a decent way to start. For example, for demon threads either the OP or the mods could ban any user from that post only (maybe with an associated karma hit, but probably not). If the commenter feels strongly about the point they are making, they can start a new post, provided they have enough karma.

Moderation Log