3 ms·
>If Youtube devs could see that significant amounts of dislikes were coming from users who hadn’t watched the video, or could identify other statistical aberrat
by matteoraso 3y ago
>If Youtube devs could see that significant amounts of dislikes were coming from users who hadn’t watched the video, or could identify other statistical aberrations, it stands to reason that such abuse would actively interfere with the legitimate functionality it was intended for and/or work against the interests of YouTube, advertisers, as well as authors and viewers.
But if you can see that these dislikes were from trolls, then you can account for that and not have the algorithm register them.
- hnarn 3y ago> then you can account for that and not have the algorithm register them It’s an unsolvable problem which is why they disabled it entirely. If you “account” for “bad” input the only consequence is that those responsible for that bad input figure out how to get it classified as good input.
- s3p 3y agoWhat? Suddenly we are in a ficitious world where there are droves of trolls disliking videos, and somehow they are sentient as to YouTube's recommendation algorithm, and they desperately want the video to disappear from everyone's recommendations, so these trolls come up with new and inventive ways of DISLIKING videos? I don't think I've ever read more made-up scenarios than on this website.
- rale00 3y agoGroups of trolls banding together to down-vote people they dislike is as old as down-vote buttons. Surely you've heard of brigading?
- sojournerc 3y agoright, I don't think the comment was disregarding bregading. That a multi-billion dollar company's algorithm is too dumb to deal with it is what astounds.
- johnfn 3y agoYou really don't think there's a single person out there who wouldn't pay 20 bucks to a farm to downvote a competitor's video? Not a single one?
- nerbert 3y agoGoogle didn't think it'd be worth it to pursue a solution that would solve this, and probably nobody internally wanted ownership of it. The second best solution is the one that doesn't cost much and solve the problem. Youtube being the only game in town, where are people gonna go anyways? Daily Motion? Post implementation KPI probably showed that traffic hasn't budged, and the problem has been solved. As far as Google is concerned, it was a rational decision.
- Kiro 3y agoYou don't even need to be famous in order to get trolls who are hating you so much that they make it their job to ruin your life. When you get above a certain size you will have organized troll armies coordinating attacks on private Discord servers.
- Spivak 3y agoDing ding ding! You basically can't allow people to give negative feedback for a thing and have that feedback mean anything (ie affect recommendations for anyone but you or show it to other users) without insincere feedback being used to hurt the reviewee. There are ways to counter this, the easiest is to not show negative reviews but count them positively, a dislike actually boosts them just like a positive review would. Not really recommended due to promoting rage bait but brigading would stop working.
- cellularmitosis 3y agoCounting them positively results in the problems which tiktok is currently facing with “rage bait” content.
- Spivak 3y agoThat would be the other consequences, yes. So can't say I recommend it.
- anonymouskimmer 3y ago> There are ways to counter this, the easiest is to not show negative reviews but count them positively, a dislike actually boosts them just like a positive review would, I think it would be better to just merely count the number of upvotes and complete, or almost complete views, for boosting purposes. The downvotes should be for tailoring feeds, whether personal, or the aggregate feeds of people with similar interests and like/dislike votes. Any specific criticism can be saved for the comments.
- Zak 3y agoYou also can't allow people to give positive feedback without insincere feedback being used to artificially promote things. If you're going to have a feedback system, you need some manner of checks and balances on it to limit insincere feedback.
- pessimizer 3y ago> It’s an unsolvable problem If it's an unsolvable problem for downvotes, it's an unsolvable problem for upvotes, too. The reason they took away downvotes is because they started to partner with the networks, to artificially boost their posts, and to deemphasize and demonetize their traditional amateur comment (inspiring a shooting.) The mainstream content gets ruthlessly downvoted because polish doesn't equal quality, and the networks (wisely) don't want their stuff distributed by a platform that allows users to mark it as bad. So Youtube took away the ability to mark content as bad. It's no more complicated than it looks.
- paulpauper 3y agoI am sure the removal had more to do with certain ideological views/content being more likely to be downvoted. The decision came soon after the critically panned Susan Wojcicki YouTube CEO 2021 Free Expression Awards. This probably has the record for the worst ratio of any video in the site's history.
- rightbyte 3y agoPeople watch videos in their idelogical niche anyways, so ideological downvoting aint that much of a problem.
- jayd16 3y agoWhy waste the bandwidth, click through rate hit, and drama over users getting shadow banned for their trolly opinions?
- scraptor 3y agoBecause users clearly care about the information it provides them, to the point of implementing third party solutions to restore it as best they can.
- valianteffort 3y agoHow would you even distinguish people who immediately realized the video was shit and so downvoted and didn't watch from trolls?
- zakki 3y agoI think we can infer that troll doesn't need time to watch video. So if a dislike came from a user with watching time is less than 10% (just as an example) it can be categorized as troll.
- mattl 3y agoI’ve definitely watched less than 10% of a lot of videos and would downvote them if it was easier on TV. Lot of videos either start with a nonsense intro/bad audio.
- Narishma 3y agoI've disliked many videos based on the first few seconds, or just the title alone if it's click-baity enough..
- avereveard 3y agoBy comparing standard deviation from every other rating's time. By comparing account voting patterns across videos. By comparing registration time and voting time. By identifying grouped clustered votes time relationship. This just on top of my head, going to bet there is going to be more to be found in aggregate.