Live data from Hacker News

Mozilla Is Crowdsourcing Research into YouTube Recommendations

foundation.mozilla.org

81–90 of 179 posts

Re: Mozilla Is Crowdsourcing Research into YouTube Recommendations

#81

Seems silly Why not just build a better (user-configurable) recommendations engine?

I don't think it's a bad suggestion, though their plan is most likely to influence people to complain to Google.

Same as politics, you want a maximum of people to be convinced by your idea/opinion/etc to make it happen faster. If you build your idea it might be great but if no one is convinced of it it won't work.

Re: Mozilla Is Crowdsourcing Research into YouTube Recommendations

#82
post #63

I'm always mind-blown reading comment threads like this regarding recommendation, with people annoyed that explicit feedback helps recommenders, or exasperated that they're recommended one item that they haven't approved while also wanting novel recommendations. 1000s of hours of videos are uploaded every minute. 5+ billion videos are watched daily. Are you not rather quite pleased that YouTube can deliver highly rel…

The complaints are due to a significant decline in the quality of recommendations. It’s like that scene from 2001 where HAL is slowly lobotomized by Dave.

Re: Mozilla Is Crowdsourcing Research into YouTube Recommendations

#83

What's impressive to me is that YouTube Recommendation are so ... stale. All it recommends to me is content from the same 3 categories, and the same videos over and over and over (you would think if I didn't click on the suggestion in the past 6 months it would adjust). Now the topics it recommends to me are non-political, but I can very well see that if you fall down one of more extremist rabbit holes thats all you…

I use YouTube infrequently. Usually for binging random content like comedy sketches or speedrunning. No matter what I'm on, there's always a few videos in my recommendations that make no sense and look like trash. Not generically, but specific videos that ALWAYS appear. One of them is "I PAID FIVE BASSISTS ON FIVR TO PLAY AN IMPOSSIBLE BASSLINE" ... Why is this so bad? What are your hundreds of data scientists doing?

> One of them is "I PAID FIVE BASSISTS ON FIVR TO PLAY AN IMPOSSIBLE BASSLINE"

Slap like right now

Re: Mozilla Is Crowdsourcing Research into YouTube Recommendations

#84
post #47
post #8

Earlier quoted context omitted.

Recommendation bubbles are partly privacy matter -- without the ability to track users, there would be no recommendation bubbles. It may sound like a far-fetched possibility -- how to maintain privacy of logged in users? But I figure it is doable from technical standpoint; the key question would be how to transform the financial part of the operation to keep Youtube afloat. At this point, even the awareness of recomm…

You can recommend just based on content. I hardly ever open the youtube homepage. And often open links in incognito mode. Still get to see all these recommendations next to a video. I don't see how you could not have a "recommendation bubble" as long as you have recommendations. Even if they were to be 100% manual it's still a bubble of some kind. Not only that, you effectively have these bubbles outside the internet…

Another name for recommendation bubble is sub-culture. Google has decided that there should be only one and the results are just as inane as a high school clique taking over the school newspaper.

Re: Mozilla Is Crowdsourcing Research into YouTube Recommendations

#85
post #80
post #13

Semi related: there's a browser plugin[1] available to "de-mainstream" Youtube. It removes from recommendations the well known channels of the big name media brands[2]. You can add more to the list. -- [1] https://demainstream.com/ [2] https://github.com/miscavage/De-Mainstream-YouTube-Extension...

This works pretty well (I use it) because it removes recommendations and search results (which are really recommendations) that are artificially inserted/ranked at the top every single time. That said I don't think it's what Mozilla and many commenters here are looking for, they're looking for banning content they disagree with (i.e. censorship).

> they're looking for banning content they disagree with (i.e. censorship).

That's an extremely unfair read of the situation. The extension doesn't even ban anything. It is a research project attempting to understand why the YouTube algorithm recommends the videos it does. Whether or not you think Plandemic (a nonsense conspiracy theory video) should be banned or not you can still see research into the YouTube algorithm as a worthwhile endeavour.

Re: Mozilla Is Crowdsourcing Research into YouTube Recommendations

#86
post #29

A few months ago YouTube was flighting a feature I liked a lot. The homepage would group my interest into named categories "Cute dog videos", "Strangeloop", "Debussy", "PyData", "Algebraic Topology" and so on. I really liked that because I could choose which deep end to sink into. Not just that it would also serve as a reminder on what I had been interested in, but have become too distracted to remember. They seem to…

I'm pretty sure they're still testing that feature. I see it show up occasionally.

Interestingly, I think there are a couple of variants for the number of videos to display on the home page, too.

Re: Mozilla Is Crowdsourcing Research into YouTube Recommendations

#87
post #63

I'm always mind-blown reading comment threads like this regarding recommendation, with people annoyed that explicit feedback helps recommenders, or exasperated that they're recommended one item that they haven't approved while also wanting novel recommendations. 1000s of hours of videos are uploaded every minute. 5+ billion videos are watched daily. Are you not rather quite pleased that YouTube can deliver highly rel…

The complaints are due to a significant decline in the quality of recommendations. It’s like that scene from 2001 where HAL is slowly lobotomized by Dave.

Since when, do you think? I personally think that they're better than ever. On the other hand, I think they moved to more deep learning approaches over the last 3-5 years (based on their research output at least) so perhaps recommendations are perceivably different since then

Re: Mozilla Is Crowdsourcing Research into YouTube Recommendations

#88
post #73

Earlier quoted context omitted.

If they were only recommending comedy from the same artist you may then have the same issue as another comment, which is that recommendations aren't novel enough. I'd take your point if they only recommended bass videos, but presumably it's just one recommendation in a list of many, with the point being to explore your preferences while also satisfying your demands

The problem everyone is complaining about is that they used to recommend a good mix of content that was actually relevant. Like if you were watching a bunch of comedy videos by one comic it would recommend more of them but also another comic that you would probably enjoy. I don’t know how they did it, maybe it was just based on what other people with similar demographics and interests watched. If it was then they pro…

If they used to recommend a good mix of content, what do you feel is the issue with recommendations now? Not enough variety? Not enough relevance?

Re: Mozilla Is Crowdsourcing Research into YouTube Recommendations

#89

What's impressive to me is that YouTube Recommendation are so ... stale. All it recommends to me is content from the same 3 categories, and the same videos over and over and over (you would think if I didn't click on the suggestion in the past 6 months it would adjust). Now the topics it recommends to me are non-political, but I can very well see that if you fall down one of more extremist rabbit holes thats all you…

That's my experience as well. It will only recommend videos that I assume are popular globally, and old videos from channel I've watched in the past (including videos I've already watched), until I watch enough of something new and it suddenly switches to only recommending that. There's gotta be some middle ground...

I don't understand how the recommendation algorithm seems so brittle, given that advertising is their core business, and that I am manually doing so much of the work by subscribing to channels I enjoy. A simple feed of recent updates from channels I subscribe to would be an improvement; in fact I think I mostly find videos to watch from Twitter.

Re: Mozilla Is Crowdsourcing Research into YouTube Recommendations

#90
post #29

A few months ago YouTube was flighting a feature I liked a lot. The homepage would group my interest into named categories "Cute dog videos", "Strangeloop", "Debussy", "PyData", "Algebraic Topology" and so on. I really liked that because I could choose which deep end to sink into. Not just that it would also serve as a reminder on what I had been interested in, but have become too distracted to remember. They seem to…

I'm pretty sure they're still testing that feature. I see it show up occasionally. Interestingly, I think there are a couple of variants for the number of videos to display on the home page, too.

How can I get myself on the better side of their A/B test ?
Post reply on HN