Hacker News Reader

Comment by caconym_ | Hacker News Reader

I'm with you. It baffles me how much people claim to care about a metric that seems only to be legitimately useful in the edgiest of edge cases, and has literally never been of any use to me. And, maybe I'm being unfair, but many of these people seem like exactly the sort of people who'd get involved in an ideologically-motivated downvote brigade. Ironically, all the signals I do trust to tell me which videos are worth watching are telling me that this one, even if it contains some grain(s) of truth, is probably heavily biased and not worth my time.

The fairest critical take I've seen (not that I totally agree with it, but I respect his experience as a creator) is probably Linus' "rant": https://www.youtube.com/watch?v=HN3ojSzyIG8

View on HN

> It baffles me how much people claim to care about a metric that seems only to be legitimately useful in the edgiest of edge cases, and has literally never been of any use to me.

Different use cases for YouTube, presumably? If you're listening to music or watching well-established content creators, then dislikes are basically irrelevant. But if you're looking for a "how to" tutorial, dislikes and comments are the only way to quickly sort the good from the bad.

I think comments are less useful since the channel owner can arbitrarily delete and filter them (correct me if I'm wrong). So part of the outrage is that dislikes were the last semi-reliable metric left to check for garbage/troll videos.

check out flat earth videos... they remove all comments that are critical and now with the dislike counter gone i guess the earth is flat.

> But if you're looking for a "how to" tutorial, dislikes and comments are the only way to quickly sort the good from the bad.

This is the "edge case" (not the best term for it, but whatever) I'm talking about, but I still don't really buy it. There are a lot of signals for how worthwhile a video is going to be, and frankly I'd rank likes/dislikes as the least useful one, mostly because they provide no context and aren't normalized to the video's view count.

If we see an epidemic of amateur electricians getting zapped, etcetera, I suppose YouTube will have some egg on its face. But somehow I don't think that's going to happen.

>There are a lot of signals for how worthwhile a video is going to be, and frankly I'd rank likes/dislikes as the least useful one, mostly because they provide no context and aren't normalized to the video's view count.

What are those other signals to help filter quality? Because in my experience, the dislike-to-likes ratio is the #1 statistical signal. It beats the #likes/#views ratio which is far weaker. Comments are not a better signal because (1) they can be deleted by the channel and (2) they take more work to scroll through than just looking at the public dislikes number.

It doesn't even matter that dislikes don't have "context" (do you mean "downvote reasons"?). To me, the context is the video itself.

If I come across a plumbing how-to and it has more dislikes than likes, it's a junk video. MKBHD is another youtuber that explains the usefulness of dislikes counts: https://www.youtube.com/watch?v=CaaJyRvvaq8

EDIT add: also note in the MKBHD video the comments from other famous youtube creators that they also depend on public dislike stats when they themselves consume videos.

The primary signals I use to judge the worth of a video are 1) its prominence in search results, and 2) the video itself.

This is because dislike counts are in my experience extremely unreliable—they may be elevated for stupid and irrelevant reasons on a really great video, or they may not be significantly elevated on a video that is very much not worth my time. Anecdotally (though no more so than any other account in these comments, AFAICT), I've found this to be true even on videos with relatively objective subject matter.

>The primary signals I use to judge the worth of a video are 1) its prominence in search results,

This is not as good a signal because for many long-tail videos, the topmost video in the search results has a bad dislike-to-like ratio.

>and 2) the video itself.

This is not efficient use of time for the viewer because the point of a quick statistic is to avoid watching the video in the first place.

>This is because dislike counts are in my experience extremely unreliable—they may be elevated for stupid and irrelevant reasons on a really great video,

Totally opposite experience. For thousands of videos I've consumed that are not-politics and not-music such as tutorials, the bad dislike ratios were the most reliable indicators Youtube had for trash videos.

To be clear, I'm talking about dislike ratios and not absolute counts. Maybe that's the source of the disagreement? E.g. if I see a video with hundreds or thousands of dislikes (absolute counts), it's not a problem unless the ratio is out of whack.

The problem with this discussion is that (presumably) neither of us will ever have the data to back up our position. We can only disagree based on how we ourselves use YouTube, and maybe how we've observed friends and family using it.

So what is the point? Unlike a lot of people I've discussed this with, I can at least respect your position as being based in some kind of believable and relatively well-developed use case. But nothing we've said here has even touched on the (alleged) negative externalities of visible downvote counts that (allegedly) motivated this change in the first place. And, again, we just don't have the data to understand the tradeoffs there.

I entered this discussion because I really, honestly don't understand why so many people are so angry about this. And, despite your efforts, I still don't. I have no trouble believing that these legitimately bad, long-tail, high downvote count videos exist, but I still see them as exactly what my original comment characterized them as: an edge case.

I guess I'm just looking for a bombshell that will make me understand, because that's how obvious the downvote count stans make this issue out to be. You aren't wrong just because you don't have such a bombshell, but you aren't going to convince me you're right just by telling me that the way I use YouTube is wrong.

>The problem with this discussion is that (presumably) neither of us will ever have the data to back up our position. [...] I have no trouble believing that these legitimately bad, long-tail, high downvote count videos exist, but I still see them as exactly what my original comment characterized them as: an edge case.

Because the public dislike counts are now gone, examples are hard to find. In any case, here's a screen shot of a popular video from The Verge that's not long-tail. The deep link showing 1200-dislikes-vs-839-likes: https://youtu.be/WuunLhXcUo4?t=15

That's an example of a high "dislikes ratio" alerting the viewer to a trash video. Lots of bad videos with corresponding dislikes on Youtube are not an edge case. I encounter them every day (e.g. product reviews, DIY how-to, etc)

>, but you aren't going to convince me you're right just by telling me that the way I use YouTube is wrong.

If you're happy using Youtube without considering dislikes, we're not saying you're wrong. The point is that your dismissal of public dislikes is not relevant to how _others_ depend on it to avoid wasting time. The examples of famous youtube creators also looking at dislike counts to save time is very telling. Even though they themselves suffer from dislikes on their own channel, they still depend on dislike counts when consuming others' videos. The youtube creators are also some of the most sophisticated viewers of Youtube content and their usage of dislikes ratios matches the reality of many in this thread who use the metrics in the same way to filter out bad videos.

This is all well and good, but—again—what you aren't addressing is the (claimed) negative effects of visible downvote counts, mainly (as I understand it) spurious downvote brigades against relatively small/vulnerable creators. These use cases you and others (claim to) care about so much must be weighed against the (claimed) damage done by continuing to support them, so it's not just a question of live and let live.

When downvote brigades are brought up, the first response is always something along the lines of "all engagement is good" and "downvotes aren't a negative signal in the algorithm", the thrust being that downvotes don't actually hurt creators and they're just being a bunch of crybabies. But it is, frankly, totally incoherent to claim this and then turn around and also claim that downvotes are an important signal for deciding which videos to not watch. I hope you can see the contradiction there.

I point this out, and the next claim they trot out is "downvote brigades are rare to the point of being insignificant". To which I say, well, prove it. Prove it, show me the data, and make the case to me that saving X man-hours of wasted time watching unworthy how-to videos is worth crushing Y creators' nascent careers.

They care because none of this seemed to matter until the White House and then the big media, who is being tagged as authoritative despite getting a lot wrong, were ratioed consistently.

People asked why disliking POTUS matters now, when it didn't before.

People asked why big media gets to fact check everyone, despite obvious, daily bias and error, and dislikes are hidden for them and this helps everyone how exactly?

People noticed some creators are widely disliked consistently, yet enjoy consistent promotion. They noticed money seems to matter.

Creators who generate many ratioed videos are different from those who don't, or who have a mix of some kind too.

All of these were brought up in discussion, brought to me by normies basically unable to buy into the, "you do not need that info because..." line they were given.

And now that they are talking about it, topic of interest, they are wondering about other things, like why their Facebook keeps removing what they thought were private exchanges...

Your bomb shell is watching people lose trust and see less value in all of this than was true a year ago.

I don't think videos on politicized topics getting "ratioed" is a very good signal of anything in today's political climate. Many, many users are willing to downvote content featuring people and/or organizations they've been told are their enemies, without engaging with the content at all.

In fact, this sort of thing is IMO one of the best arguments for removing downvote counts.

Beyond that, I frankly don't think people should have trust in "platforms" telling them what to think, so if this change wakes up some of the "sheeple" I can only see that as a good thing. I certainly think that taking away a tool that allows people to trivially express their shallow hatred of a thing can only improve the state of discourse and critical thought in society.

This is all trying to tell people what to think!

Pages of discussion on how data should be interpreted.

All the while ignoring the big players who did not like the fact that they are widely disliked.

As for taking away a tool...

POTUS has a very low approval rating. The dislikes follow, as they should.

That same POTUS getting after long overdue policy with majority public support would very likely see a different outcome.

People disliking big media and government sell jobs? SHOCKER!

Of course they did, and of course it leads to questions which, of course led to eliminating the source of those questions.

Finally, "I don't think people should..." Full stop. We need to realize those other people have agency and work from there.

I watch how-to videos a lot, and I think dislikes are not that useful. In my case either it's either A) a common topic, and I just go by total view count and watch the top 3-5 videos or B) it's a super niche topic and I watch all 1-5 videos available.

It feels like people are imagining some fantasy scenario where there are hundreds of similar videos with not that many views but very different vote count patternsm...

so your telling me its fine to leave only likes for a video showing you how to make a DIY air conditioner ? using a bucket full of dry ice and a fan?

100% this will end up killing someone.

Content likely to end up killing someone should be flagged and removed, not community-downvoted.

That particular example may have been exaggerated or extreme, but I think the point still stands.

Someone just commented this "check out flat earth videos... they remove all comments that are critical and now with the dislike counter gone i guess the earth is flat.".

Indeed. In this case for example it should not be flagged, but down-voted by the community.

I honestly couldn't tell whether that flat earth comment was being ironic or not, but I think it's an interesting case study either way. Is a visible YouTube downvote count really the only thing standing between a meaningful number of people and belief that the Earth is flat?

I don't really buy it.

> Is a visible YouTube downvote count really the only thing standing between a meaningful number of people and belief that the Earth is flat?

I do not think so either, to be honest, but perhaps there may be something to it.

Perhaps not just different use cases but different user types as well.

By that I mean for example, I knew of many people who were outraged and dismayed when RSS feeds started to disappear from websites. They belonged to a class of individual who prefers things to be categorized as granularly as possible. That kind of person is highly upset when their categorizing toolset is downsized.

But other people never bother with those tools, don't care if they're eliminated, and frankly find it hard to sympathize with their hypercategorizing comrades.

I'll disagree. It seems you and GP have been lucky enough to never stumble on videos with #dislikes > #likes, for whatever (usually well-deserved) reasons.

I'd seen some; wishing I could unsee those.

Reply on news.ycombinator.com