What is meta? ×
Meta Stack Exchange is where users like you discuss bugs, features, and support issues that affect the software powering all 134 Stack Exchange communities.

My votes yesterday triggered fraud monitor for another user and removed all my votes for him over the last 9 months. I'm certainly not a sock puppet, and the votes are all deserved.

Guessing the trigger was something like discussed in this question previously.

As well as keeping up to date with recent posts, I often pick a random page of questions by a user or a topic and vote up or down where relevant - which is what I did yesterday, as I have done many times in the past.

Just feeling a bit miffed - the user definitely deserved all my up (and down) votes on those questions, but because I triggered something he has lost out!

Update: I have trawled through my history and it looks like this has happened to two other individuals I have upvoted in the past. There was no flagging to me that my behaviour was out of norms, and I wouldn't have noticed this most recent one if the user concerned hadn't pinged me to ask what had happened. I gather from some of the other questions on this topic that there is no rollback, as the deleted votes are deleted from history - bit of an assume guilty until proven innocent issue, as even if proven innocent, the sentence still remains. (I know, I took the analogy a bit far, but still!)

share|improve this question
2  
I think the issue is that ALL post votes for the user has been removed, not just the votes from the day you were looking at their user page. –  Ian Ringrose Jul 22 '11 at 14:23
    
@IanRingrose: Yep - absolutely - every vote I ever gave him –  Rory Alsop Jul 22 '11 at 14:26
    
Voting "fraud" is something else (think, sockpuppets). Although I see other folks have been using that term WRT this as well... Probably because they're both described in the same blog post, and it has "fraud" in the title. But this is the ol' "serial voting" script - users are supposed to be voting on posts, not users, so if one user votes on posts from another too much, this kicks in and negates it. Trust and believe, if you got caught trying to defraud the system you'd be losing more than a few votes. –  Shogging through the snow Jul 22 '11 at 17:11
1  
@Shog9 - but because I went over the limit on voting yesterday it has negated ALL my votes I have ever given him. –  Rory Alsop Jul 22 '11 at 17:40

3 Answers 3

up vote 6 down vote accepted

I'm not sure that this is really a "false positive". The way you describe your workflow:

I often pick a random page of questions by a user or a topic and vote up or down where relevant - which is what I did yesterday, as I have done many times in the past.

seems to be exactly what the automated vote fraud scripts are looking for and designed to rollback. That is an anomalous voting pattern in any definition of the word.

Now, of course, you could always make a compelling argument that your votes were correct and should not have been removed, but there's no way of putting that logic into the script. You can't have it look at the reputation of the voting user, because sometimes even high-rep users might let their tempers get the best of them and go on a voting spree. We don't really want to allow that.

I agree this is an "edge case", and I sometimes worry that I'll trigger the automated vote removal as well. I frequently downvote all of the bad questions asked by a user when they come to Meta and complain about another one of their [bad] questions being closed. The hope is that this will get them one step closer to a permanent question ban. But it's all in vain if those votes are rolled back, so I try to be careful when I do this.

I'm not sure what else to recommend in this case, other than considering a change to your workflow. If you really feel strongly about the votes you cast, you could email the team at team@stackoverflow.com and ask them to investigate and [possibly?] replace your votes.

More information:

share|improve this answer
    
I'm not sure the votes could be replaced by the team. I think they're really gone from the system. –  ChrisF Jul 22 '11 at 9:42
    
@ChrisF: Yeah, I suspect you're probably right. But I have this image of the devs as all-powerful overloads commanding an entire army of unicorns that can do basically anything they want. –  Cody Gray Jul 22 '11 at 9:43
3  
+1 Already decided to change my workflow - I read a couple of those links you posted, and it seems you are right. I just feel a bit bad for the individual concerned as he has a lot of really good answers. –  Rory Alsop Jul 22 '11 at 9:45
3  
Sometimes it is because someone just answered something really well, so I have gone back to look at their other answers. Sometimes it is because I have looked at everything on the front page and want to do a random dive in to a page I have not really looked at. –  Rory Alsop Jul 22 '11 at 9:52
9  
Interesting. I don't often follow this pattern much, so I doubt that I've ever triggered this, but I have read multiple posts by a particular person before, following links from the Users/info page. If I find one question or answer by a particular user interesting, the odds are good I'll find others interesting too - so why shouldn't I do that? This assumption of fraud seems a bit like excluding a popular author from the best sellers list because lots of readers bought the whole trilogy at once. BTW - how often does Jon Skeet lose votes this way? I'm sure it's a drop in the ocean, but... –  Steve314 Jul 22 '11 at 10:56
8  
-1 because I don't see this as fraud. The filter is measuring the metric it's designed to measure, but the metric doesn't reliably indicate fraud. An anomaly doesn't reliably prove an abuse, and therefore IMO there need to be checks and balances. At the very least that means remembering what votes are cancelled, notifying the affected voter, and allowing for appeals with the possibility of a rollback. –  Steve314 Jul 22 '11 at 11:02

I'm not sure how practical this would be (I don't know how common the problem is, so I have no idea if it's overkill), but here's an idea.

When the fraud-detection script gets a hit...

  1. Create a page for the potential fraud.
  2. Include links to all the problem votes.
  3. Include a vote-as-fraud/as-not-fraud widget with a current counter.
  4. Add a link in the "review" page, and notify the affected voter with a link.
  5. Maybe allow that voter to include his defence in an answer-like box - not sure if there's much point to this, but maybe.
  6. When the vote counter exceeds certain positive/negative limits (and with minimum and maximum time limits), make a permanent decision based on that vote. If the maximum time is exceeded with net zero votes, probably err on the side of caution (though that's probably ambiguous, depending on your viewpoint).

IOW, the robot detects the anomoly, but the community gets to decide if it indicates genuine fraud.

share|improve this answer
2  
Interesting idea, but probably too much work for the number of times false positives show up. –  ChrisF Jul 22 '11 at 11:23
1  
I like this - it does put it back in the community's court, not an automated tool. –  Rory Alsop Jul 22 '11 at 11:24
2  
This really misses the point that voting is anonymous. Not an acceptable solution. –  Cody Gray Jul 22 '11 at 11:42
    
I don't understand who would have access to the voting information in your proposal. Currently, voting is completely confidential except that community moderators know when a user has cast a “suspicious voting pattern”, but even then the details of the pattern aren't revealed. –  Gilles Jul 22 '11 at 11:44
1  
@Cody - voting isn't anonymous if the voter chooses to self-identify, e.g. by a comment. So maybe a simple modification would be that this page only gets created if the voter chooses to appeal, having received the initial notification? Clearly the data exists somewhere, or else the robot couldn't detect the anomolous pattern in the first place. OTOH, if the whole thing's overkill... (I've calmed down a bit since my "at the very least" comment) –  Steve314 Jul 22 '11 at 12:06
    
@Steve314: In general, we'd prefer to discourage that type of self-identification. "+1 blah blah" comments are considered noise. And the whole point of having it done automatically by a script is that no human users ever have to look at that data. Yes, the computer has to store votes, but they don't have to be exposed to anyone, including the developers. As I understand, that's how the system works now. This would be a huge step backwards for vote anonymity, which is a very important principle around here. Your choice to appeal shouldn't publicize votes against my post(s). –  Cody Gray Jul 22 '11 at 12:09
1  
@Cody - OK for most, but for the last sentence, the recipient of a vote didn't make that vote - he's not responsible for it, and shouldn't be affected one way or the other if the voter is identified. That ignores puppets, of course, but would a genuine puppet choose to appeal? And should we worry if he did? Extra note - a lot of people get upset if there's downvotes without a "-1 blah blah" comment explaining why - is that really noise? –  Steve314 Jul 22 '11 at 12:25

The use of the term "fraud" makes it sound like people are doing really nasty things, which @rory clearly was not trying to do.

But I would agree with Steve314 (and now Rory) that the workflow he was using may well introduce a subtle bias into reputation calculation, by rewarding "cliques" of people (either friends or those with similar interests or whatever) who tend to innocently visit the user pages of others in the clique, follow their links and upvote when they like them.

I'd suggest that the rep earned by upvotes from people who got to the questions/answer from somewhere other than a user page are more reliable.

Do the scripts track referrers to know if that is the case? (Or would you have to shoot us if you told us? :)

share|improve this answer
1  
Yes, in this particular case, the word "fraud" is relatively deceptive. But there's no actual message displayed that accuses him of voter fraud; this is just something he's concluded on his own, looking at existing topics that discuss things that can trigger the vote fraud script. Most cases that it picks up are indeed fraudulent. I mention in my answer that I think this is an "edge case", but that I'm not sure how we can fix it without breaking either our ability to detect and remove votes that are actually fraudulent, or revealing voting history which is supposed to be anonymous. –  Cody Gray Jul 22 '11 at 16:41
    
I don't see why it matters how the voter got to the question, though. In either of the common cases that the vote fraud script is intended to detect (sock-puppet voting or enraged, vindictive user voting), that user would find the questions/answers to upvote from the user's page. –  Cody Gray Jul 22 '11 at 16:42
    
I wasn't taking the terminology personally - the script is designed to stop vote fraud, and I agree with it in principle, I'm just feeling guilty that one of my techniques for finding good answers has penalised somebody else through no fault of their own. –  Rory Alsop Jul 22 '11 at 17:43

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .