It's annoying clicking some youtube link only to find it was something you really don't want more of, and then youtube proceed to recommend much more of it, even though you barely watched any of it.
kinda forces me to watch something else that is new but at least somewhat wanted in an attempt to make the algorithm look at the new shiny i'm interested in and forget the previous one.
i find it funny though when amazon still make recommendations to me based on a purchase in 1998. yes, Jeff, i still want to buy wrestling VHS tapes...
> amazon still make recommendations to me based on a purchase in 1998. yes, Jeff, i still want to buy wrestling VHS tapes...
It’s funny you mention that.
For many years I would get recommendations for Latino lgbt books. Weird as it’s not really my topic. I found a way to look at how they generate recommendations, and it turns out that I bought a book about a gay Chicano growing up for a college class in 1998 or something.
It's frustrating that nowadays YouTube seems like it can only keep 3 or 4 types of videos that you like in its head. I feel like I'm watching the same thing over and over whenever I click on recommendations. I totally agree with you.
Exactly! Sometimes my recommendations are good, with long form tech and history content I actually want to watch. But click on one Top Gear clip and that's all out the window, and I have to actively recall what these channels were called and try to build them back into my recs.
There was a time a couple years ago I watched one episode of Arthur for nostalgia purposes and it completely obliterated my recommendations. I scrolled for hundreds of videos and didn't see a single non-arthur one. It took months for the site to become usable again.
I continually get suggested videos that are all already watched. I can't think of the last time I've gotten video suggestions that don't already have a red line at the bottom. At least it has stopped queueing up the same 3 videogame walkthrough videos I always click out of the minute they start.
The fewer categories they use, the less manpower and computer power they spend on classifying and censoring content. They serve millions upon millions of queries a day with millisecond response times, and they don't get paid for providing value to you, so that's what we get. I hate what the internet is these days.
What if it's from a channel I like but don't want to watch that one video...will clicking on this reduce suggestions from this channel, will it do nothing, will it unsubscribe me from that channel(as has happened in the past), ...?
Nobody knows, I only know I don't want to play this game.
It doesn't in my experience. I've been telling it I don't want to see: news, politics, music, or reaction videos for months. It still puts them there.
Last week it decided that I wanted to see rap videos and filled my feed with them. I don't listen to rap. I've never watched a rap video. I've told it I'm not interested in every single video, but it still puts them there.
I’ve been trying it for a few months. I dutifully clicked away every preposterously stupid clickbait title or thumbnail and topic I didn’t care about. It had no noticeable effect.
I wish someone would make an extension that makes this a simple hover+keypress instead of multiple clicks. As it stands right now it is too much of a hassle to manually click 2 times per video. But if you do not then youtube will continue to plaster that video on your homescreen for weeks.
Doesn't work well at all. There's a popular garbage news channel here in Brazil.
Every week I have to do that, it never works.
I never click on the videos, so no idea why they think I'm interested.
YouTube's personalization recommendation is a nightmare. Just the other day, YouTube played "Billie eilish -copycat" for me literally after every other song.
I pretty much only YouTube in Incognito mode these days. Everytime I forget, YouTube manages to annoy me so much within half an hour that I switch to Incognito again.
Yes, I always hesitate to mark YT videos as unwanted because often it's from a channel I otherwise like a lot and I have no idea how YT will behave after that. Considering the site is known for even auto-cancelling subscriptions sometimes it feels like a UX minefield.
This will kind of clean up your current recommendations but it doesn't reset the recommendation algorithm like it used to. The algorithm still retains memory of your past behaviour.
For example if I consistently listen to 3 unrelated songs [A B C] on youtube together, the algorithm will regularly recommend them to me (because of my unique behaviour, not because they're similar). If I reset my history and then listen to song A, then B and C get immediately recommended even though they aren't similar to A and they don't exist in my listening history which means that the information about my listening/browsing habits is still there in the recommendation model.
It’s better if you click that stuff, it causes the algorithm to recommend stuff you don’t care about, which means less YouTube holes to get sucked into
It's crazy just how much I go out of my way to avoid things that "personalize" my experience.
I don't use online music services, I discover music on various platforms and and download mp3s and keep a local library.
I avoid Youtube at it's defaults, I use 3rd party apps and VLC to do most of my watching, other than my subscriptions I tend to skim the Home page very rarely.
I do not use Netflix or other streaming services, I try to hunt down DVDs/Blu-Rays and prefer ripping them for my personal library.
My only problem is exclusives, as a fan of The Witcher series, I do feel like I am missing out, but if I feel a really strong urge I can always borrow an account from a friend, create a temp profile, watch the series and delete it.
Their convenience features just add more inconvenience to me.
I don't know how common this is, but I've gotten into the habit of creating "personas"-- different users that have different habits and are interested in different things, yet all me.
It's insane that I feel it helpful to take on "multiple personalities", but there it is.
Part of it is that these algorithms are fairly one-track. They can mix it up a bit, but it's always too much of one thing and too little of another. They can't truly comport with the reality that someone can have multiple interests and tastes.
It infuriates me that we basically had this as a feature, lost it, and have to sort of reimplement it with an almost illicit feel.
I can recall in the mod-late 1990s when Mainstream America was starting to go online for the first time, everyone was training their kids "never give out your real name." From that simple bit of stranger-danger paranoia, we built a lot of communities as psuedonymous by default-- your AIM "screen mame" was rarely your given name, you could have different usernames on each forum, your email address probably referenced your favourite sports team or anime character.
This inherently constrains aggressive "passive" personalization. Without an obvious canonical identity, you don't want to try to cross-profile too aggressively, because user "hakfoo" on site A may well be different from user "hakfoo" at site B, and you had to assume that any ID you tracked was limited or transient: when you go off to college or apply for a job, you're probably not going to want to be slapping "PonyGirl1987" on your resume.
I wonder if it's that the algorithms are limited to being one-track or if they're overoptimized to being one-track though. I assume they have a profile somewhere that looks like "12% coin collecting, 31% travel to Paraguay, 9% 2004 Ford Focus Repair, ..." and then they offset that data with what content produces the best revenue/engagement/metric of the week. In the process, many secondary interests simply get demoted to the pont you see nothing but tire change tutorials.
What you're describing is the "real" Internet, which peaked somewhere around 2007. Everything was pushed about as far as it could be in that direction.
The problem was (almost) no one was making money off it. Online advertising, as it was originally done, was a joke. Most of the early internet was based on ideals of quality, community, and freedom. None of these make much money.
So people started hunting for what DID make money, and they discovered data harvesting and targeted advertising. And like a cancer, that unholy pair devoured most of the web. Information really WAS power, like all those breathless articles and hacker manifestos in the 90s said. Power and MONEY. And the new masters of this realm have dedicated themselves to taking as much of YOUR information as possible. They need it, like Elizabeth Báthory would have needed the blood of virgins if she was an actual vampire and not just insane.
another solution is to use 2 instances of chrome or firefox with different --user-data-dir which can run in parallel . It's actually very convenient for checking up stuff without 'messing' the main setup
Same, but I did this long before the current tracking nightmare. That's the whole FUN of "cyberspace": you can create and destroy identities at will, which means you can experiment. Even better, you can live out multiple lives simultaneously to some degree, like multi-threading code.
Realistically, on the modern net you need to: Use a VPN, kill cookies and ads aggressively, have multiple accounts (or avoid logging on at all in some cases), and never, ever use 2FA except with financial institutions. That's the bare minimum. No point in just complaining that tech companies are evil all the time. The smart organism adapts to its environment, and uses the tools at hand.
Yes buts he’s code switching in order to fool an automated panopticon that’s integrated with our most advanced communications technology. That’s a bit unique.
I sometimes want to rewatch a funny scene, but the fear of having to sit through a 2min ad on Hulu after I press the back button outweighs the few seconds of joy I'd get from that scene. Nor is it worth the money to upgrade.
This is the default experience of vlc, mpv, and kodi. You have to do the surveillance-navigating work when obtaining content, but then when you're relaxing you're free to relax.
It's amazing that in 2021 piracy is still superior to all the paid products in terms of quality and user experience. They've had over a decade to figure this out, and they just don't care.
I like recommendation features in principle. I don’t want an unfiltered feed of stuff.
It would just be nice if they were configurable, understandable and actually worked for me. Some systems’ “We’re showing you this because ...” is an important first step, but it needs to go way, way, further.
Big companies probably don’t do it because they’re afraid of overwhelming a user, disclosing too much about their own algorithm and the things they know (“We’re showing you this because someone you hung out with on Instagram just before you met”).
Thus it’s probably up to open source projects again to make algorithms that actually work for people, not against them.
Are there any such projects already? Are there projects trying to dissect common proprietary recommendation systems? They’re some of the most mysterious influences on current society.
My hobby project https://linklonk.com (use invitation code 'hn') is a content recommendation system that is built around these principles:
- Control: what you see is based on your explicit upvotes & downvotes, not what you happened to click on.
- Transparency: you see content from feeds that posted content you upvoted and from other users who upvoted what you upvoted, but did that before you. For example, when you upvote a link it will tell you that "You will get more content from 2 users that also liked it and from 3 feeds that posted it". And when you see a recommendation from other users you can see what likes in common do you have with those users (“We’re showing you this because ...”).
- Fairness: the amount of attention every user and feed gets from you depends on how useful their past recommendations have been. Ie, the higher their signal-to-noise ratio has been for you - the more prominently their other upvoted items will be in your "feed".
Give it a try and let me know what you think. I'd like to do a "Show HN" post for my project soon and your feedback would help me prepare for it.
If you like to read a bit more here are my announcements with discussions:
There is one problem really. These companies are not charging money for their services. It's the same problem with other mega-corps. They can afford short and medium-term losses to kill other players entering the field. This is the root cause of so many problems in capitalism.
There are of course other challenges preventing people and corporations from entering this field but at this point I don't even think there's hope. They have not only destroyed everyone else in the private sector, but they partnered with governments and political entities in such a way that their continued rule is basically cemented in place. You won't get the recommendations that you know would be good for you because they're not trying to do something good for you, they're trying to extract more value from you.
If you spend a minute too long looking at the menu for a movie it will not only auto play, but for the foreseeable future Netflix will assume you’re watching it, want to finish watching it and watch similar stuff.
More than most other services Netflix doesn’t trust your ratings or your lists when making up recommendations.
Fundamentally, tools should serve their masters. When it's the other way around, it's time to find a different tool.
And yes, they had to get rid of ratings, because some people's feelings got hurt because they were being rated so low. Can't have that pesky objective reality rearing its ugly head.
It feels like Netflix knows what I like and throws things I don’t like at me on purpose, just to see whether I’ll change my mind. (So I unsubscribed)
The discovery is specifically what makes Youtube, Netflix or Spotify sticky beyond just content hosting, and even at suggestions, pirate websites are better at it.
You can disable autoplaying of next episodes in a series that you're watching, and you can disable autoplaying of previews, but I don't see any way to disable autoplaying of movies and shows in general when you're looking at their info pages.
I don't log in to anything google related, I have youtube/google stuff containerized and when I close my browser window it clears all my history and cookies, etc. I surf with ublock on and max privacy settings in the browser.
I use YouTube via RSS. My quality of life is incredible.
In the novel, "Feed", one of the characters has a hobby of searching for all sorts of random things, in order to confuse the algorithms. She would add absurd items to her shopping cart, without buying any of them.
Readit News
kinda forces me to watch something else that is new but at least somewhat wanted in an attempt to make the algorithm look at the new shiny i'm interested in and forget the previous one.
i find it funny though when amazon still make recommendations to me based on a purchase in 1998. yes, Jeff, i still want to buy wrestling VHS tapes...
It’s funny you mention that.
For many years I would get recommendations for Latino lgbt books. Weird as it’s not really my topic. I found a way to look at how they generate recommendations, and it turns out that I bought a book about a gay Chicano growing up for a college class in 1998 or something.
There was a time a couple years ago I watched one episode of Arthur for nostalgia purposes and it completely obliterated my recommendations. I scrolled for hundreds of videos and didn't see a single non-arthur one. It took months for the site to become usable again.
Seems to work reasonably well for me.
Nobody knows, I only know I don't want to play this game.
Last week it decided that I wanted to see rap videos and filled my feed with them. I don't listen to rap. I've never watched a rap video. I've told it I'm not interested in every single video, but it still puts them there.
I pretty much only YouTube in Incognito mode these days. Everytime I forget, YouTube manages to annoy me so much within half an hour that I switch to Incognito again.
For example if I consistently listen to 3 unrelated songs [A B C] on youtube together, the algorithm will regularly recommend them to me (because of my unique behaviour, not because they're similar). If I reset my history and then listen to song A, then B and C get immediately recommended even though they aren't similar to A and they don't exist in my listening history which means that the information about my listening/browsing habits is still there in the recommendation model.
Deleted Comment
I don't use online music services, I discover music on various platforms and and download mp3s and keep a local library.
I avoid Youtube at it's defaults, I use 3rd party apps and VLC to do most of my watching, other than my subscriptions I tend to skim the Home page very rarely.
I do not use Netflix or other streaming services, I try to hunt down DVDs/Blu-Rays and prefer ripping them for my personal library.
My only problem is exclusives, as a fan of The Witcher series, I do feel like I am missing out, but if I feel a really strong urge I can always borrow an account from a friend, create a temp profile, watch the series and delete it.
Their convenience features just add more inconvenience to me.
It's insane that I feel it helpful to take on "multiple personalities", but there it is.
Part of it is that these algorithms are fairly one-track. They can mix it up a bit, but it's always too much of one thing and too little of another. They can't truly comport with the reality that someone can have multiple interests and tastes.
I can recall in the mod-late 1990s when Mainstream America was starting to go online for the first time, everyone was training their kids "never give out your real name." From that simple bit of stranger-danger paranoia, we built a lot of communities as psuedonymous by default-- your AIM "screen mame" was rarely your given name, you could have different usernames on each forum, your email address probably referenced your favourite sports team or anime character.
This inherently constrains aggressive "passive" personalization. Without an obvious canonical identity, you don't want to try to cross-profile too aggressively, because user "hakfoo" on site A may well be different from user "hakfoo" at site B, and you had to assume that any ID you tracked was limited or transient: when you go off to college or apply for a job, you're probably not going to want to be slapping "PonyGirl1987" on your resume.
I wonder if it's that the algorithms are limited to being one-track or if they're overoptimized to being one-track though. I assume they have a profile somewhere that looks like "12% coin collecting, 31% travel to Paraguay, 9% 2004 Ford Focus Repair, ..." and then they offset that data with what content produces the best revenue/engagement/metric of the week. In the process, many secondary interests simply get demoted to the pont you see nothing but tire change tutorials.
The problem was (almost) no one was making money off it. Online advertising, as it was originally done, was a joke. Most of the early internet was based on ideals of quality, community, and freedom. None of these make much money.
So people started hunting for what DID make money, and they discovered data harvesting and targeted advertising. And like a cancer, that unholy pair devoured most of the web. Information really WAS power, like all those breathless articles and hacker manifestos in the 90s said. Power and MONEY. And the new masters of this realm have dedicated themselves to taking as much of YOUR information as possible. They need it, like Elizabeth Báthory would have needed the blood of virgins if she was an actual vampire and not just insane.
https://www.huffpost.com/entry/data-and-goliath-digital-surv...
https://support.mozilla.org/en-US/kb/containers
https://support.mozilla.org/en-US/kb/profile-manager-create-...
Realistically, on the modern net you need to: Use a VPN, kill cookies and ads aggressively, have multiple accounts (or avoid logging on at all in some cases), and never, ever use 2FA except with financial institutions. That's the bare minimum. No point in just complaining that tech companies are evil all the time. The smart organism adapts to its environment, and uses the tools at hand.
I eventually realized I could skip around and enjoy the parts I liked. Freely. Without that feeling of being watched.
Flashback.
This is exhausting.
It would just be nice if they were configurable, understandable and actually worked for me. Some systems’ “We’re showing you this because ...” is an important first step, but it needs to go way, way, further.
Big companies probably don’t do it because they’re afraid of overwhelming a user, disclosing too much about their own algorithm and the things they know (“We’re showing you this because someone you hung out with on Instagram just before you met”).
Thus it’s probably up to open source projects again to make algorithms that actually work for people, not against them.
Are there any such projects already? Are there projects trying to dissect common proprietary recommendation systems? They’re some of the most mysterious influences on current society.
- Control: what you see is based on your explicit upvotes & downvotes, not what you happened to click on.
- Transparency: you see content from feeds that posted content you upvoted and from other users who upvoted what you upvoted, but did that before you. For example, when you upvote a link it will tell you that "You will get more content from 2 users that also liked it and from 3 feeds that posted it". And when you see a recommendation from other users you can see what likes in common do you have with those users (“We’re showing you this because ...”).
- Fairness: the amount of attention every user and feed gets from you depends on how useful their past recommendations have been. Ie, the higher their signal-to-noise ratio has been for you - the more prominently their other upvoted items will be in your "feed".
Give it a try and let me know what you think. I'd like to do a "Show HN" post for my project soon and your feedback would help me prepare for it.
If you like to read a bit more here are my announcements with discussions:
- https://tildes.net/~tech/u7f/linklonk_a_link_aggregator_with...
- https://www.reddit.com/r/RedditAlternatives/comments/mpqnpl/...
There are of course other challenges preventing people and corporations from entering this field but at this point I don't even think there's hope. They have not only destroyed everyone else in the private sector, but they partnered with governments and political entities in such a way that their continued rule is basically cemented in place. You won't get the recommendations that you know would be good for you because they're not trying to do something good for you, they're trying to extract more value from you.
If you spend a minute too long looking at the menu for a movie it will not only auto play, but for the foreseeable future Netflix will assume you’re watching it, want to finish watching it and watch similar stuff.
More than most other services Netflix doesn’t trust your ratings or your lists when making up recommendations.
And yes, they had to get rid of ratings, because some people's feelings got hurt because they were being rated so low. Can't have that pesky objective reality rearing its ugly head.
The discovery is specifically what makes Youtube, Netflix or Spotify sticky beyond just content hosting, and even at suggestions, pirate websites are better at it.
I use YouTube via RSS. My quality of life is incredible.
You could also just put the channel URL into your RSS/Atom reader and it might convert it automatically.
The book, written in 2002, was very prescient.