LW Update 2018-06-11 – Vulcan Refactor, Karma Overhaul, Colored Links, Moderation Log

by Raemon3 min read12th Jun 201834 comments

32

LW Moderation
Personal Blog

I had written about the new Strong Upvotes feature over in this post, but a) I wanted to include it in an official update post, b) we've made a number of small tweaks in the past month that I wanted to summarize.

VulcanJS Refactor

Much of our progress in the past month was under-the-hood. During the Strong Upvotes rework, we needed to make a lot of changes to the VulcanJS core voting functionality. This was making it increasingly hard to keep tabs on which changes we've made to VulcanJS.

So we have now refactored the LW codebase into 3 different repositories. You can view all of them here, and the three subsections below:

This means that people contributing to the LW Open Source will need to clone all three branches into an upper level folder. (Instructions on how to do this have been added to the Lesswrong2 readme file)

Most of the time, developers should only need to make changes to the Lesswrong2 repo, and changes to the VulcanJs repos should only be made as a last resort.

Small/Big Upvotes

This was discussed in the Strong Upvotes post. On desktop computers, you can click-and-hold on the upvote button to do a larger upvote. On mobile, click twice. (After we've had a chance to see how this has played out on both device types, we may tweak the exact implementation. This is a pretty experimental change and we appreciate all feedback about it)

Karma power has been rescaled to make sense in the small/big upvote scheme, with most longterm users have a "small" upvote power of 2, and a big upvote power ranging from 4-12.

After some period of experimentation and verifying that we've gotten the right balance of karma power and inflation, we'll be retroactively applying it to old votes to ensure that old posts don't have weirdly low karma scores. The plan here is to make all old, regular upvotes into Small Upvotes (i.e. karma of 1, 2 or 3, depending on your current karma). You can then turn old upvotes into Strong Upvotes if you think it's important.

Hover-Over for vote count

Rob Bensinger and Wei_Dai had both raised interesting-but-opposite points about the value of being able to see how many votes of each type had been cast. For now, we decided to include information on how many votes, total, a given post or comment has received, but not how many upvotes/downvotes/smallvotes/bigvotes.

My hope is that this will give enough info that you can tell roughly how much engagement a post has gotten, without overly-emphasizing the "aaah people on my team are getting downvoted" feeling that sometimes happens.

You can now hover over the karma-score of a comment to see the number of votes a post or comment has received.

Green Links

The original LW2.0 launch had links being much, much subtler than old LW. This was part of an overall concern about how much the web is constantly demanding your attention, which can disrupt reading flow. We stuck with this despite some initial complaints to give people time to adjust and see how it felt.

After 9 months of trying it out, we decided to go back to colored links. A few points that put it over the edge for us:

  • When scanning a page quickly, it can be hard to see where the links are
  • Having no color on a post at all reduced the amount of visual cue about what site you're on.
  • Most importantly (in my opinion), LessWrong is a site built upon reference to older posts, and a common way for newcomers to get acquainted with our old content is to click around in a TVtropes-esque tabsplosion. Making this easier seemed important.

Moderation Log

I don't think I ever formally announced this, because there were some bugs I still hadn't ironed out (the listed user on a deleted comment is actually the user that deleted it), but since it's been several months I wanted to at least note that the basics of the moderation log Said Achmiz had recommended has been in place, over at https://lesswrong.com/moderation.

I'll try to fix that bug soon but it's a long list of things that need doing.

Bug/Small Fixes and Open Source

  • Link posts in the search bar now properly open their LessWrong post instead of directly linking to their target link (so you can read comments and commentary)
  • Thanks to James Lamine for updating the readme file with some of the newer documentation
  • Congrats to Forrest Wolf for making his first open source contribution, fixing a small bug on the notifications page.
  • Thanks to crybx for fixing a bug with the RSS feed's author tag.

32

34 comments, sorted by Highlighting new comments since Today at 10:43 PM
New Comment

Question/comment about the moderation log feature:

If the following are all true:

  1. … the content of a deleted comment is always hidden;
  2. … the name of the person who posted the deleted comment is not shown;
  3. … the reason for the deletion is not given;

… then:

  1. There is hardly any meaningful difference between “public” and “not public” deletion; but more importantly…
  2. There is not actually any useful information provided by the moderation log. (“The mods deleted someone’s comment on thread so-and-so” is not useful.)

I am pleased to see the moderation log feature implemented, but dismayed that it does not actually seem to provide any transparency! I worry that there has been some misunderstanding concerning the question of just what the purpose of such a feature ought to be. (Surely no one thinks that a log that simply consists of entries to the effect that “The mods deleted a comment of unspecified content posted by an unspecified user, for an unspecified reason” serves any purpose…?)

I see that the missing name of the offending commenter is a known bug—but fixing that bug alone does not really go very far toward addressing my concerns. So my question is: what further changes, if any, are planned for the moderation log feature?

The lack-of-username of the actual user is just a bug, pure and simple, that I kept forgetting to fix.

The reason is provided (the column to the right). Often the reason is empty (we get lots of spam, and eugin clones, and occasional people who posted a comment twice which mods often quietly remove), and we generally err on the side of one-click deleting those just to get them out of the way.

Some changes I think we'd probably want to make (but this is off the cuff) are:

  • fix the aforementioned bug
  • maybe create easy one-click options to provide the reason of "spam", "eugin", and "double-post" in particular, since those are 99+% of deleted comments, and generally uncontroversial.
  • create a view that shows deleted comments by non-mods (the purpose here being to check if anyone is actually using deletion from their personal-moderation-settings, which I'd expect to be the biggest concern that the moderation log is trying to address). If combined with previous option, maybe ability to also filter out common reasons. [This change seems to be the most important one to me, since it cuts through the noise and lets you actually see the most potentially concerning deletions]

I might also add a small "this post has X deleted comments" at the bottom of a post, although this might also require something that excludes spam and eugins, since those would be common enough to make the notification near-useless.

There might be broader changes that involve more of an overhaul of the system (the moderation log as-is was something I through together quickly when I realized it could be done fairly easily, as opposed to something we thought carefully about and optimized)

The reason is provided (the column to the right). Often the reason is empty (we get lots of spam, and eugin clones, and occasional people who posted a comment twice which mods often quietly remove), and we generally err on the side of one-click deleting those just to get them out of the way.

Well, this is what I see right now if I go to the mod log:

If there’s a field for the reason, but you don’t put anything in there, then that doesn’t count as “the reason is provided”!

Once again I must emphasize that the point of this feature is transparency. You know that you deleted a comment for a good and uncontroversial reason. But you would know that even without the moderation log. We (non-mods) do not know that, and even with the moderation log—as it stands now—we still do not know that!

As to your list of prospective changes, I certainly support them. (Although I note that a jury-rigged approach to the second one is simply to have a Sticky or a TextEdit file or something that you keep open, that has some common reasons, for you to copy&paste into the “reason” field.)

That said, I am curious to know whether the idea of showing the text of the deleted comment is entirely off the table; and if so, why? It seems to me that this would make the feature tremendously more useful… but I don’t recall that the matter was addressed to my satisfaction, when last the moderation log was discussed.

I fixed and deployed the "user is the same as deleted by user" issue.

Excellent, thank you!

I'm all for being able to agree that spam is spam. Potentially with the "phone number" or other contact details deleted so that it's not searchable.

I'd be against the "Eugine" comments being visible because the user has a habit of reposting comments word for word after they have been banned for vote manipulation. To let them be recorded - even in a moderation log is to not delete them fully. With these comments - the username is indicative too.

Well I would like to see what this Eugine fellow has to say. What benefit do you gain from locking his words behind more than the trivial inconveniences required to tend to the garden? Reddit with its undeletion mirrors seems to do just fine.

LessWrong 2.0 is necessary because LessWrong 1.0 was killed, on purpose, by one troll with a vendetta who made it sufficiently un-fun to play in that everyone left, and created armies of sockpuppets after being banned. We are long past the point of taking anything he has to say seriously.

I notice that a lot of the comments in this thread that defend having Eugine's comments invisible have been downvoted. That could well be "on their merits" -- there are certainly good arguments for not making deleted comments completely invisible -- but if some moderator is feeling zealous, it might be worth looking for signs of Eugine-socks that might hitherto have been missed...

...yup, that looks pretty likely.

Sorry, that was my bad. It looks like I broke the thing that reversed all of Eugine's votes everytime we banned him. I fixed it now, and the votes on this and other threads should soon be restored to normal. I did a bit of analysis, and it seems that he made a total of about 80 votes or so, so a good chunk, but nothing that overwhelmingly skewed the discussion outside of this thread (since those votes were distributed over a very large number of comments).

Do you think it would be detrimental to LWs fun if I am able to dig up what he's saying? Do you think it's more about punishing him than protecting the Gurkenglas attack vector from his ideas?

Do you think it would be detrimental to LWs fun if I am able to dig up what he's saying?

Yes. Because moderation is exhausting, and every time we have to rehash it costs moderator time and attention, leading to things ranging from "less attention for more important things" to "moderator burnout." This is in fact essentially my whole point in the first place.

But why would we have to rehash it?

If you have the posts visible in the moderation log—where everyone can see that they are, indeed, the same old Eugine comments, reposted over and over—then there’s no reason at all why we’d have to rehash anything. Everyone can scroll through the log, see the same lame comments posted over and over, and feel secure in the knowledge that all that’s happening here is that a troll is being rightly kept out of the garden. No one will need to comment on this, and no one will.

Whereas if they’re hidden, then folks begin, quite rightly, to wonder just what it is that you have to hide so thoroughly.

"No one will need to comment on this, and no one will." -- bets please, lizardman constant, etc etc.

Tradeoffs. Do the mods have to deal with more BS because they show the text and a handful of people are like "but reeeeaaaallly? I got something good from that one comment!" or because they hide the text and a handful of people are like "but reeeeaaaallly? What if you're acting in bad faith?" I assume that's what's going on here.

Or maybe it's just wishful thinking. "What do we have to do to make Eugine just go away forever and we never have to talk about it again?" ==> disappear everything and hope

Tradeoffs. Do the mods have to deal with more BS because they show the text and a handful of people are like "but reeeeaaaallly? I got something good from that one comment!"

This, basically. A lot of individual Eugin comments look fine. It's only when you know the entire history of him that it's clear how important it is that he banned and stay banned. He optimizes his comments for soaking up as much time and attention as possible.

We delete about 10 Eugin comments around once a week. It's a pain. Everything that makes it more of a pain is detracting from our ability to do things like:

  • give friendly responses to newcomers who look like they have something to offer but could use some help understanding the site culture
  • resolve disputes between longterm members
  • actually code stuff (since several of the mods are also developers)

A lot of individual Eugin comments look fine. It’s only when you know the entire history of him that it’s clear how important it is that be banned and stay banned.

But I’m not suggesting that you shouldn’t ban him. I agree that you should ban him. I agree that you should delete his comments. I have zero quarrel with this.

What I still don’t see is how any of that implies that the text of his deleted comments—much less that of deleted comments in general—shouldn’t be visible in the moderation log.

Everything that makes it more of a pain …

But why would having the moderation log show the text of deleted comments, make it more of a pain to delete Eugin’s comments? I’m not trying to be dense here, but I’m afraid I just don’t see any connection…

Here’s a question, that might make me better understand your view on this. Is the issue here that you, for some reason, specifically object to having Eugin’s comments be displayed at all, in any way? And if the answer is “yes”, then would it be reasonable to suppose that you would have no objections to displaying the text of deleted comments in general, but having a special “Eugin exception” (i.e., where the text of the deleted comment would normally be, there might instead be some text in the vein of “REDACTED, because this was a comment by a Eugin.”)?

(Similarly, perhaps there could be—as I think I might’ve suggested in the past—a “doxxing exception”, and a “the contents of this comment violated U.S. law exception”, etc.)

I don't have strong opinions on whether Eugine's comments should or shouldn't be visible in the deletion log, but here (I think) is the best argument for making them not be: It's about incentives. Eugine wants LW to be his soapbox; in so far as his actions still have any motivation to them beyond mere malice, his goal is to propagate his opinions and punish those with conflicting opinions; the best hope of making him go away is for him to get nothing from posting to LW. If his comments' text is preserved, then that gives him an incentive to keep posting them.

(I fear that in fact there is nothing left but malice, and the mere knowledge that he's wasting moderators' time is enough for him. But I hope he hasn't gone so far down the path from "reasonable human being" to "entity of pure malice" for that to keep him at it indefinitely.)

[Note on spelling: definitely "Eugine" rather than "Eugin", though I believe neither is his real name.]

Alright, I suppose that’s an argument. Thank you.

In that case, my question about the “Eugine exception” approach stands.

seems reasonable to me. (including also tags in comment text if necessary)

I’m afraid this really makes very little sense.

If the mods want to enforce “no one comments on topic X”, they can easily do so.

The two alleged alternatives you list are not at all symmetric. In the former case, there’s simply a disagreement with a moderator decision; the response to that is “yep, you’re entitled to your opinion, but this is the way it is”, and further discussion after that can simply be disallowed. If anyone doesn’t like that, they can leave; but everyone knows where everyone stands.

Whereas in the latter case, you’re eroding the trust of your user base, in a quiet way, a way that will slowly corrode the administration’s relationship with the users, lead to interminable, vague arguments (because there’s no clear, public, specific decisions and actions to talk about—just suspicion, innuendo, suppositions, and conjecture), and take up far more energy and use up far more good will than simply a firm but public stance.

You can go talk to them not on lw.

banned for vote manipulation

Banned for what now??

Edit: Oh, the “they” in your sentence referred to the user, not to the comments. (How confusing!) Well, in that case, I don’t understand your logic. Why does the fact that a troll is reposting comments after being banned, mean that those comments shouldn’t be visible in the comment log? Help me out, here; I’m not following your reasoning.

They are unwelcome to make any impact on the site. And at this point they are still going to be unwelcome five or ten years from now. Comments are daily. And deleted daily.

I'd be happy if the log didn't even reflect the existence of the deleting of the comments.

You’re still not making any connection between those two things.

E was banned. That decision is locked in.

Consequences of being banned include - you don't get to post here and people don't get to read your comments. Even in a moderation log.

… and people don’t get to read your comments …

So, to be clear: you propose to punish the entirety of the Less Wrong commentariat for the transgression of one person? That seems extreme, not to mention unproductive.

To be clear. I'm talking about only the E comments. The rest should be in a log.

And I think of it as protecting LW from a persistent bad actor. Despite multiple attempts to ask for them to join mediation or talk about what the problem is or even do anything other than repost the same text of a comment.

Once again, I need absolutely no convincing of the fact that Eugine’s lifetime ban is well-deserved. I certainly never did nor ever would suggest that any attempt at “rehabilitation” should be made.

My concerns are exclusively about the good of the site and its non-banned users, not about any “rights” of Eugine, or leniency, or charity, etc., toward him (as nothing remotely like that is warranted).

What I am questioning is whether hiding Eugine’s deleted comments is good for me (for instance; and for the rest of the commentariat). I am not at all convinced that it is. This is what I’d like to see a cogent argument for.

That said, it seems like “show deleted comments, but make certain extremely rare exceptions—such as for Eugine” might be the least controversial solution (if not 100% ideal from all perspectives—but perhaps “100% ideal” is an unrealistic goal).

It's not obvious to me whether access to Eugine's deleted comments is a net benefit to the LW readership.

Reading something has a cost (in time, attention, etc.) as well as whatever benefits it brings, and a lot of the things we might read are rubbish; so we seek out various sorts of probably-better-than-average writing. Removing something from a given corpus of writing may well be beneficial, if the thing removed is of notably lower quality than the rest of the corpus and isn't required in order to make sense of the rest of it. (This applies even if the thing, in absolute terms, isn't so bad.) Or if, in absolute terms, it's not interesting enough to be worth the trouble of reading it.

It is plausible to me that Eugine's mod-deleted comments are in fact of notably lower quality than the rest of the mod-deleted comments, if only because they are so damn repetititititititive.

"Quality" here means "whatever it is that one might be looking for in reading through deleted LW comments". That might be much the same as quality of ordinary comments, if you're reading the deleted ones for fear of missing something good. (In that case, my guess is that Eugine's comments are better than average for deleted comments -- at least, if there were some mechanism for collapsing the vast numbers of dupes. But I doubt many people will be reading the deletion log on the off chance of finding hidden gems.) It might be amount of information about moderator decisions. (Eugine's deleted comments are very low in such information, because they all reflect a single decision to delete all his comments.) Etc.

Let's suppose arguendo that the ability to read Eugine's deleted comments is on average of benefit. I'd like to note that it wouldn't follow from that that they should be made accessible, because e.g. the deterrent value of having it be known that if you behave like Eugine then you are liable to be maximally excluded from the LW community might be sufficient to outweigh that benefit. (If someone commits a serious crime and is imprisoned for it, this greatly reduces others' access to conversation with them, to their ideas, etc. You could call this "punishing the entirety of the rest of the world for the transgression of the one person", but that seems to me an unhelpful way to look at it.)

To recap this comment and an earlier one of mine, I suggest three reasons why it may be reasonable for Eugine's comments to be permanently destroyed. (1) To reduce his incentive to make them. (2) To deter future Eugines. (3) Because including them may be of negative net benefit because they're particularly uninteresting even among deleted comments on LW. On the other side, I guess we have (a) the merits of a policy of never deleting comments permanently, for the sake of transparency, and (b) the possibility that to some people Eugine's comments might be very interesting or valuable. It's not obvious to me how (1,2,3) weigh up against (a,b), but it does seem obvious to me that it isn't obvious that (a,b) massively outweigh (1,2,3). Do you agree?

(I ask the last question because one impression I get from your comments on this is of incredulity, as if you find it baffling that anyone would think it makes sense to delete Eugine's comments permanently. I'm not sure whether I'm imagining that, nor whether if I'm not the incredulity is real rather than adopted; if it's real, then I hope to have made it less baffling that some people want Eugine's comments nuked.)

whether hiding Eugine’s deleted comments is good for me

yeah. I can see how you might want to make that decision for yourself. In which case you can probably try to get in touch with this person off platform.

I give Mods the permission to sometimes make these decisions for the users.

I don’t understand why you keep missing my point, but I have run out of ways to explain it. Perhaps someone else can step in for me.

It was my memory from the discussion where the idea of having the moderation log was born, was that including the post text per default makes sense.

In a forum where there's one trusted moderator team that does all the moderation decision I don't think it's an issue when deleted posts are completely invisible. LW2 however isn't such a forum but one where individual users are supposed to set moderation policies for their own posts and moderate according to those policies. A moderation log that includes the text of the deleted post would provide a way to understand the defacto moderation policy of every user who moderates his own posts.

This will likely lead to some discussion about moderation that take some attention but when the idea is to experiment with different moderation styles I don't think those moderation discussions are useless.

What will happen to individuals' karma when old votes are reweighted? The Right Thing, in some sense, is to go through vote by vote in chronological order, recalculating karma scores as it happens, and reweight each vote according to the user's recalculated karma at that point. I don't know whether that's painfully expensive either in computer time or in developer time. If it is, it seems like the obvious alternatives have consistency problems...