Edit: Please stop upvoting me. I'm beyond where I was when this whole mess started. Thanks, and I feel really stupid for the way I felt when I wrote this first paragraph now.

 

First of all, I'm sorry for this thread. I largely expect it to get downvoted, but I am considering leaving LW over this, and I thought I would make one more effort to actually get some answers before I did. Because I'm pretty confused. I've never seen something like this happen on LW before. I've been downvoted faster than any troll I've ever seen on here.

Three days ago, I got into a thread over a point that was admittedly pretty pedantic and minor. For two days, I got mildly downvoted (average of about 0.5 downvotes per comment), which was no big deal. But a day later, all of my comments in that thread started getting wildly downvoted. I lost over 100 karma in an hour, and have lost another 100 since. This was entirely on comments, since I have no main level posts that would easily kill my karma that quickly. Interestingly, comments that had stabilized at positive karma also got downvoted, even though I had made them, in some cases, a week before the downvoting began.

Specific examples:
Was at 0 karma for approximately 24 hours. Now at -5.
Was somewhere from +1 to +3 (not sure) for five days. Now at -7.
Was +2 for five days. Now at -7.
Was at +2 for five days. Now at -3.
Was at +1 for six days. Now at -1.
Was at +6 for a week. Now at +2.

Was at -1 or -2 for about five days. Now at -6. Note: This started at negative karma. Although that fluctuated highly around the time of posting. It was anywhere from -4 to +1, I believe.

Initially, when this happened, I thought (p = 0.75) I was being karmassassinated. Now, I am not entirely sure what to think (hence this thread).

I postulate a few hypotheses:
1. I am being karmassassinated.
2. All of the downvotes are legitimate downvotes, caused largely by halo effects - ie. when people see a largely downvoted comment by me, they go to my profile to read my other comments and downvote those as well.
3. The downvotes on my previously upvoted comments were expressions of a desire to give more than one downvote per comment to the thread that started this.

My prior, given over 100 karma lost in an hour, obviously, was karmassassination. But it would be difficult for a karmassassin to give large amounts of negative karma to a relatively few number of posts, however, given the restrictions on new accounts to downvote. (I don't think you can downvote until you have positive karma?) Given this and the consistent and steady loss of the additional 100 karma over the last two days, however, I am updating towards 2. At this point, I would assign 50% probability that it is at least some of both.

I know that fishing for karma is almost always seen as poor taste and downvoted. I don't mean to do that. I don't really care that much. (Although losing 33% of my karma was a pretty big slap in the face.) The thing I care most about, is probably my related PredictionBook prediction, which was one of the first things that actually crossed my mind. What I would like, is explanations of why others think this might happen. Am I so incorrect to have deserved over 200 individual downvotes? If the downvotes on my previously upvoted comments are legitimate, why do those deserve downvoting? I am really at a loss here.

 

Edited to fix hyperlink.

New Comment
61 comments, sorted by Click to highlight new comments since: Today at 4:43 AM

One thing you didn't mention - probably because you don't know that - is that your comments in the discussion thread you linked run contrary to a position that's very firmly established on LW: that it's only rational to donate to one charity, the one you judged to offer the maximum utility. For example, this post makes this point very forcefully, but it also runs through several of Eliezer's posts in the sequences. Your arguments in the linked thread can be seen as annoying rehashing of arguments amply covered and refuted in those posts, and that could cause people to downvote you in greater numbers than you would guess based on the comments' content.

However. Having said that, the loss in karma is very heavy, and seems to affect comments of yours that have nothing to do with that thread, and don't really deserve negative karma. Assuming your numbers are correct, I see three possibilities:

  1. You're being karmassasinated by random LW users irked by your comments in the linked thread.
  2. You're being karmassasinated by a group of LW users working in concert for whatever reason ("hey guys, he sucks - let's all go and downvote his recent comments")
  3. You're being karmassasinated by a sockpuppet network.

I find myself thinking that 3 is the most likely explanation given the evidence (that surprises me). However, I don't rate it much higher than 1.

Well, I wasn't attempting to argue that it was rational to donate to more than one charity. I don't think it is. I was trying to provide reasons someone might. But that's beside the point.

Thanks for your input though.

I don't really have much to add, but since I mildly expect this to be downvoted, I'd like to add as a data point that a) I don't mind this post though I would not like such posts to be commonplace, b) the downvotes seem suspicious to me too, c) I do not think they were fair and d) I would personally be a bit upset if this happened to me and wouldn't want to be further punished for having brought it up.

Thanks, I don't mind the karma as much as I would like to get other people's data points on what's going on. So you've given me exactly what I wanted.

My prior, given over 100 karma lost in an hour, obviously, was karmassassination. But it would be difficult for a karmassassin to give large amounts of negative karma to a relatively few number of posts, however, given the restrictions on new accounts to downvote. (I don't think you can downvote until you have positive karma?)

Hmm... could be somebody's been actively maintaining multiple accounts to keep this option open; or there could be an organised brigade of downvoters who have really taken umbrage. Both seem unlikely, but not as unlikely as many people individually deciding at the same time to downvote many of your posts. A puzzler.

The former is my guess at this point, given that I've gotten over 100 upvotes since I posted this (thanks guys). And I looked at the only one who responded to my karmassassination comment, and it looks a lot like a dummy account:

http://lesswrong.com/user/CallMeSIR/

This comment does not seem to warrant its current number of upvotes:

http://lesswrong.com/lw/8s6/video_qa_with_singularity_institute_executive/5g9o

Edit: For posterity, when I first observed the comment, it was at +14 karma.

Ah, I see how it works:

1) Make several accounts. 2) Post comments with those accounts. 3) Use those accounts to upvote your own comments. 4) Once sufficient karma is achieved, downvote others like crazy.

5) ???? 6) PROFIT!!

[-][anonymous]12y30

put a dot and two spaces instead of a ) after the numbers to make an ordered list.

1.  this is first
2.  this is second

shows up like:

  1. this is first
  2. this is second

EDIT: Doesn't look all that impressive, but it is an html numbered list.

Its sibling also has 14 undeserved upvotes, which seems strong evidence for 14 sockpuppet accounts being either used offensively on RobertLumley for some reason, or tested before bending the conversation more subtly in other threads.

Yeah, but he also has some legitimate other posts to his name, which weakens that case.

Did you notice this comment?

Looks like even the good guys maintain sockpuppets.

Edit: Or create them when need arises.

Yes, I did. I wasn't sure how to feel about it. I'm pretty sure he alone at least gave me 75 karma back.

I would strongly prefer that people didn't make sock puppets at all, ever (even to right a wrong like this).

I don't want LW to become a battleground for bands of rogue internet users, slinging hordes downvotes onto unsuspecting contrarians.

Poetics aside, LW doesn't appear to have a large enough community for sock puppetry to be ignored (as this incident shows). And having the threat of a large karma hit for statements that are even only mildly not the "accepted position" would accelerate evaporative cooling (and make the whole website a less nice place to be).

Incorrect. I up-voted no posts.

What was the point in making this account then?

And shortly after you posted that comment, my karma steadily increased about at the rate it would take someone to upvote every post I'd made. And old posts of mine kept going up and up, progressively further back in my history.

As a reference, user: sam0345 rarely receives downvotes of the quantity which user: RobertLumley has received. This is rather aberrant.

Why is this sam0345 character so heavily downvoted?

A habit of expressing his opinions without proper framing. He also occasionally seems to be an anti-liberal of the 'reversed-stupidity' variety.

Yeah, looks like sockpuppetry.

I've been on this site for several years, and I still don't know who is in charge at the deepest level (technical administration, let's say), so I don't even know who you can appeal to for help, whether IP addresses for downvotes etc are logged, and so on.

Trike Apps are the main maintainers/developers, but I believe Eliezer is involved still (and the code is open source, and they do accept contributions).

whether IP addresses for downvotes etc are logged, and so on.

I've looked at some relevant sections of the source code, but I'm unsure if I should say anything publicly.

Seems like the accounts that made each downvote would have to be stored somewhere, since anyone can go back and change their votes at any time. So it should be easy enough to mine the database for curious patterns if you knew what you were looking for.

Paul Graham at Hacker News has hinted that he has all sorts of hidden machinery in placing looking for odd voting activity. Given how big and occasionally ugly it has gotten, Reddit probably also has all sorts of inside workings going on, though I don't know how open they are about them.

Robert, has any moderator or system administrator contacted you about this and said they'll look into it? Or is this episode just going to sit there unexplained and unaddressed?

No, and I don't necessarily think they need to/should. The situation has been resolved; EY (who is the only person I'm sure could actually do something about it) has better things to do; and the only potential solution I can think of that is relatively easily implemented is banning the accounts, which isn't really long term.

We've found one dummy account, and it would be good for the community to keep an eye out for other dummy accounts, but the problem doesn't seem serious enough to me to warrant expenditure of significant resources on it.

Karmassassination is my best guess. Most of your posts seemed not to be bad, and in a couple of cases they seemed insightful.

This comment's karma really surprises me though, not because of the post quality, but because 41 people looked at and downvoted it.

This comment's karma really surprises me though, not because of the post quality, but because 41 people looked at and downvoted it.

I think a lot of people are tired of roland bring up that issue again and again. (And possibly (negative) halo effect once it got a few downvotes and replies explaining why people downvoted.)

[-][anonymous]12y20

There is such a thing as an unreasonable position (a crackpot, if you like). The difficulty is in accurately distinguishing between correct contrarians, the reasonably misinformed, and the absurd.

This being a wonderful example of the negative effects of that difficulty.

How do you know that?

...because it's at -41 karma? That many dummies would take quite a bit of work.

I see it at -2 right now... That's really strange.

Sorry, wrong link somehow. Fixed.

"but I am considering leaving LW over this"

Is the utility of LW to you so small that a mere change in your karma number makes the consideration of departure reasonable?

I don't care much about my overall karma score, but I'd be pretty irritated if all of my comments immediately got voted below the visibility threshhold.

No. I didn't really care that much about the karma. It was more about being presented with a set of evidence that LW would actually downvote me to that extent, largely without explanation. That's not really the type of community I wanted to be a part of. But fortunately, it's not the type of community that LW is, apparently. My calibration over this entire thing has been really, really confusing. I initially thought it was karmassassination, but after no one really said anything in support of me, and I got downvoted to like -10 for that comment, I figured it was legitimate voting and people were trying to tell me to stop complaining. After this thread, I've again updated.

And regardless, I would have probably still tried to finish reading the sequences. Which I'm really doing a horrible job of...

For what it's worth, I recently took what I considered an unreasonable karma hit on a particular point I made. It was the most karma I ever lost on a single post, and I'm quite sure I've been a bigger jerk in the past. So my estimate is that more people are exercising their right to ding karma these days.

For what it's worth, I agree that that comment is not worth -8 karma and have upvoted you accordingly. I probably would have done that for any karma level below -1. But some of the backlash may have been to a couple of things: 1. I do think you were wrong (I agree with wedrifid's comment) and 2. You made a comment mentioning how much you were being downvoted and were rather stubborn and argumentative about it. This (my post) is quite literally the first thread, comment, etc. that I've seen someone make about how much they've been downvoted that actually lead to upvotes.

Edit: Fixed hyperlink to wedrifid's comment.

Recommenting so you notice my edit. I didn't realize I linked to the wrong comment by wedrifid. Sorry for any confusion I may have caused.

Hmm regarding the donations to multiple charities - sorry for offtopic but I just want to post something vaguely in agreement.

I see that if the function which determines which charity is the best, may be exploitable, donating to top 5 charities or so which are substantially different (not cloning each other) would reduce the pay-off for finding an exploit in this function.

I think of it this way. If i were to make a charitable bot, would it be most effective for this bot to donate to just 1 charity that it determines based on some function? Big red bold NO because this bot will be deliberately gamed (on top of failing consistently) It's why we need Google to return >1 search result. The first result may not work for the user at all.

Adding jitter/dithering to inexact calculations to improve resulting behaviour is common practice. If you have a target-shooting AI in a game, when it's shooting at it's best prediction of the target position, it is easy to evade. If it jitters some, it may score far more hits.

edit: Matter of fact I did write a shooting AI that did just this, and it made the AI more effective. If said AI was sentient but not clever enough to understand my rationale, this AI would either justify it with faulty logic or try to get rid of the fuzziness and jitter i've added (and then this AI would ponder why it is not becoming more effective).

Humans of course would, too, implement this sort of jitter and added fuzziness to their decisionmaking they'd pick from others when growing up, usually without understanding of the rationale behind it; backing the heuristics with faulty logic or dismissing it based on the faultiness of the logic.

It's why we need Google to return >1 search result. The first result may not work for the user at all.

I dispute the analogy. With Google, you get the full benefit of finding what you're looking for as long as at least one of the results is right. The cost of multiple results is the time spent reading them, which is not at all the same as the cost of splitting a donation.

Well, the analogy is not perfect but there is a cost to clicking on the other links, too.

When you are splitting between top 5 charities - the difference between expected utility of them drown in the error of your heuristics used to pick top - so the 'cost' of splitting probably boils down to the cost of e.g. paypal processing the payment, and other entirely trivial costs (which you can avoid by e.g. randomizing between top 5)

On the topic of Karma, I can't be the only one here from the old Extropian mailing list, where, if I remember right, we had rather extensive collaborative filtering and voting schemes built into the mailing list software in the mid 90s.

It's pitiful that more than a decade after PageRank, the Web 2.0 has "advanced" to an additive thumbs up/thumbs down model on user forums. At this rate, we'll catch up to Usenet functionality in 50 years, and PageRank in a 1000 years. So much for the law of accelerating returns.

That was just a fad.

More seriously... having never used Usenet, I have no idea what functionality you're talking about. Care to give me a cliff's notes version?

As for PageRank... Wouldn't you be rewarding people for how often other people mention their comments? That doesn't seem to be a useful metric.

Wouldn't you be rewarding people for how often other people mention their comments?

Yes, and that actually seems like a really useful metric. PageRank for comments linking Sequence posts is another incredibly useful metric.

PageRank for comments linking Sequence posts is another incredibly useful metric.

So now your karma depends significantly on how well you can find a Yudkowskian citation for whatever you're arguing.

Nope. we're not a cult. Not at all.

(Or, for that matter, you karma now depends significantly on how gratuitously you spray your comment with links.)

Yes, and that actually seems like a really useful metric

No, because.... How often do you link to comments?

How often do you link to comments?

Personally, I seem to have a few instances available where I've done just that.

So now your karma depends significantly on how well you can find a Yudkowskian citation for whatever you're arguing.

I'm quite sick of this rhetorical trick, of conflating Eliezer with the Sequences in order to justify slinging the word 'cult' around. LessWrong is built on the bedrock of the Sequences. To the extent that we get anything done here, it's because we don't get bogged down in rehashing any of the million problems that the Sequences solved or dissolved. Honestly, this website would be one giant morass of interconnected arguments over definitions without these posts alone. The Sequences are valuable and should influence, guide, and control discussion on LessWrong. It's one of the ways we have of being less wrong! That the Sequences were written by Eliezer Yudkowsky is incidental to their importance on this website. Stop pretending like everything he's ever written is SIAI-cult-propaganda.

LessWrong is built on the bedrock of the Sequences.

Agreed, but is that doesn't mean that linking to them is the high watermark of Bayesian thought or whatever you're trying to measure?

You'd generally link to the sequences when you're speaking with someone who hasn't read them. So basically your interactions with new users and trolls now carry far more weight than they should.

Why would you link to the sequences when you're speaking with someone who's familiar with them? It'd be like bringing up 'we breathe oxygen' every few sentences.

Yes, the sequences are important. But that doesn't automatically make linking to them a good metric for usefulness to the community.

Stop pretending like everything he's ever written is SIAI-cult-propaganda.

Never claimed that, never will.

Personally, I seem to have a few instances available where I've done just that.

Have you linked to every single comment you think is relevant/useful/thoughtful? To 80% of them? 60?

So now your karma depends significantly on how well you can find a Yudkowskian citation for whatever you're arguing.

As far as I understand, PageRank rewards the target of the link, not the origin. So this would reward oft-cited articles.

That's what it generally does, but it doesn't look like that's what shokwave was thinking of (judging by the context).

If that was what he was saying, consider my last comment retracted. (Except for the very last sentence.)

I'd be interested in PageRank on LW.

Or, if not that, just the most linked to posts. (It'd be cool if one of the Trike Apps people could run a regex (e.g. "lesswrong.com/[^ ]*") over the database (reddit_data_comment[key=data] and reddit_data_article[key=article], I believe) and publish a .txt dump of that somewhere.)

I'll be in the TrikeApps office about a week from now; I'll do my best to remember this and have something workable ready to offer to them; can't promise they'll be excited about data-mining LessWrong though.

Thanks! (I'm not getting my hopes up for it.)

I've knocked out something quickly. I've got no idea how fast it will be over the ~250000 comments (there are probably some performance improvements by replacing "for ... in cursor" with a paged retrieve).

I believe that that will only keep public posts (so no drafts or deleted posts), I'm not so sure about the comments though (I'm not sure if comments on deleted articles are kept or not, or if there is such a thing as a "private" comment that I'm not filtering properly).

That script is a "best case" situation, since it records the origin along with the target of each link (and the date/karma too). If that data was to be published, I'll do try some analysis (and maybe even a proper article!).

For Usenet, you could filter posts based on scoring of posts based using regexps on the various fields of a message - body, subject, sender, etc. See the following for some details:

http://www.slrn.org/docs/slrn-FAQ-4.html

I meant PageRank as just one example of Collaborative Filtering, where Amazon and NetFlix collaborative recommendations would be another, and probably more relevant. I think we could "like" someone on the email list, and their "likes" would count transitively to us, and the likes of those they like, etc.