YouTube’s ‘dislike’ barely works according to new study on recommendations
1–10 of 437 posts
Re: YouTube’s ‘dislike’ barely works according to new study on recommendations
#2This article is a discussion of a study Mozilla published ... without a dateline, though the Wayback Machine suggests it probably appeared 20 Sept. 2022 (today): https://foundation.mozilla.org/en/youtube/user-controls/>
That was previously submitted to HN though there seems to be no activity presently: https://news.ycombinator.com/item?id=32910321>
Re: YouTube’s ‘dislike’ barely works according to new study on recommendations
#3It’s been a while since I’ve logged into YouTube but my impression is that the experience including recommendations is worse if you are logged in.
Re: YouTube’s ‘dislike’ barely works according to new study on recommendations
#4It’s been a while since I’ve logged into YouTube but my impression is that the experience including recommendations is worse if you are logged in.
Best YT experience is to clear cookies then open a lot of YT links from HN, but when our human nature drives us watching a stupid video or two from YT recommendation then after that session do clearing cookies and open YT links from HN again.
Re: YouTube’s ‘dislike’ barely works according to new study on recommendations
#5YouTube doesn't seem to understand my preferences, at all. It's far, far too biased toward what you recently watched. It basically doesn't work and certainly doesn't "learn" anything.
Re: YouTube’s ‘dislike’ barely works according to new study on recommendations
#6YouTube doesn't seem to understand my preferences, at all. It's far, far too biased toward what you recently watched. It basically doesn't work and certainly doesn't "learn" anything.
I feel like every recommendation service works this way - YouTube, Amazon, Spotify, etc. Whatever the last thing you looked at becomes your entire life's mission according to these systems. When I think about all the buzz words around machine learning, ai, and so on, it's completely incompatible with the reality I see. What exactly are all these engineering teams doing just to end up with this result?
Amazon might be the worst example though. Buy something like a bike rack for your vehicle and Amazon will recommend nothing but other bike racks... as though I need dozens of them for one car.
Re: YouTube’s ‘dislike’ barely works according to new study on recommendations
#7YouTube doesn't seem to understand my preferences, at all. It's far, far too biased toward what you recently watched. It basically doesn't work and certainly doesn't "learn" anything.
I feel like every recommendation service works this way - YouTube, Amazon, Spotify, etc. Whatever the last thing you looked at becomes your entire life's mission according to these systems. When I think about all the buzz words around machine learning, ai, and so on, it's completely incompatible with the reality I see. What exactly are all these engineering teams doing just to end up with this result? Amazon might be…
> Buy something like a bike rack for your vehicle and Amazon will recommend nothing but other bike racks... as though I need dozens
Perhaps they beta tested only with Jeff Bezos who had this exact quandary.
Re: YouTube’s ‘dislike’ barely works according to new study on recommendations
#8It’s been a while since I’ve logged into YouTube but my impression is that the experience including recommendations is worse if you are logged in.
Just don't try the YouTube search and scroll down by more than six results, it will make you scream inside.
Re: YouTube’s ‘dislike’ barely works according to new study on recommendations
#9It’s been a while since I’ve logged into YouTube but my impression is that the experience including recommendations is worse if you are logged in.
That was my experience years ago when I last did so. Logging in offered no visible upsides and multiple downs.
I learned not to log in and use search wherever possible.
Now I use Invidious (with JS disabled and/or uMatrix blocking YouTube functionality in the browser as an additional nudge), or youtube-dl / ytdl / mpv to play video directly. Usually audio-only for talkpieces.
I'll also disable recommendations and comments portions of the website, for additional bonus (via Stylus).
Re: YouTube’s ‘dislike’ barely works according to new study on recommendations
#10This article is a discussion of a study Mozilla published ... without a dateline, though the Wayback Machine suggests it probably appeared 20 Sept. 2022 (today): https://foundation.mozilla.org/en/youtube/user-controls/ > That was previously submitted to HN though there seems to be no activity presently: https://news.ycombinator.com/item?id=32910321 >
The linked report [0] is dated today and uses data from the RegretsReporter extension [1] and additional surveys from December 2021, which is basically when youtube removed the dislike counter [2].
It's both suprising and not that suprising to see quantitatively how little disliking or telling youtube to not reccomend channels or content has an impact on the algorithm. They also rightfully point out how much of a pain it is to try to provide negative feedback to youtube's reccomendation system. It's always been a chore to remove the hot new trending clickbait from the reccomendations.
I sadly doubt that this report will change much, but it's good to finally have some statistical validation of what I've frustrating subjectively experienced.
[0]: https://foundation.mozilla.org/en/research/library/user-cont... [1]: https://addons.mozilla.org/en-GB/firefox/addon/regretsreport... [2]: https://blog.youtube/news-and-events/update-to-youtube/