If it were up to my opinion. the short answer is yes. if the topics/genres that I were interested in, i.e. I searched for, I clicked, gets recommended more.
For a privacy perspective it might be a bad thing depending how it was implemented, it’s quite unlikely that the algorithm is is done externally, or locally on a users machine, it would likely be from the server and stored there too. unless we have something like a peertube client that could sync the preference data between devices.
either that, or we’ll have to encrypt it. if we so care much about privacy. but decentralization definitely helps.
Now on the topic of addiction is a whole another discussion, it may vary for user to user. I’m not an expert in this but for me I can control what I want to do fairly well. I tend to scroll through the recommendations at leisure, when I want to look for useful information like guides, tutorials, advice on a skill i’d use the search bar, categories, or try sift through the recommendations.
a while ago I was mindlessly scrolling and quickly made aware of it, I despised the feeling and avoided doing the same thing. other might be still stuck in a loop. i’m… not sure.
the recommendations from algorithm are the ones that had “gems” aka information that would helped me tremendously that I wasn’t specifically looking for and higher quality videos. It’s like a double edged sword, depending on the user’s intent, or… control of their intent.
that is definitely a problem to some people, so we should give them an option to make it more adhd friendly.
I’d like to say in defense of algorithms if it were something I knew what it was doing. I’d be in favour of. it’s more of a discoverability function, like what’s poppin on peertube y’know? I haven’t found anything good for a while, it’s very rare and like mostly found on accident or via youtube. maybe it’s just that we haven’t seen any massive adoption here, but I don’t really know, what do you think?
How to discover content on any platform federated or not is a challenge.
on mastodon it’s pretty easy. Look at any mastodon user, look at who they boost. See anything you like and give the poster a follow. Look at their boosts and repeat
Unfortunately, it’s also how various platforms have produced competing bubbles. Users can take a deep dive down both positive interests like cooking or exercise, but they can just as easily be driven towards video after video of rage porn. It’s a tradeoff.
I think this is happening, but not with mastodon and their chronological timeline.
I think it would actually be desirable if the personalization was more transparent and customizable…
Like… imagine if in the settings I could select a set of tags and give each tag a “weight” so the ones I’m more interested in show more often in my feed than the ones I’m not… and have an option so that whenever I click on a link the tags assotiated to it get a tiny percentual boost in my preferences table (under certain min/max limits, and perhaps affected by “frecency”).
This way I could have control over those settings, edit them, delete them or let them change dinamically if I wish to, while always being able to opt out completelly from them changing whenever I feel I reached a balance I like.
I’m not sure what you mean by most of this post.
Peertube has an algorithm. If you see the most upvoted videos, then that’s an algorithm. If you see the most upvoted videos from that month, then that’s an algorithm.
If you’re looking for ‘an algorithm for good recommendations’ the best out there is Reddit/ Lemmy, which simply asks people to upvote, and tags the posts, so you select, e.g. ‘dogs’, then people who like dogs vote.
It would be nice to have something like Lemmy explorer but for peertube. I find it impossible to search for stuff. Maybe what im searching for doesnt exist.
you could index the videos in a self hosted independant search engine. intelligent ranking (without rich relevant metadata) is a more involved project. if you mean personalized recommendations… then i dont really know. it has a nasty stigma attached to it.
yes
deleted by creator
