I’ve noticed that there are a few communities that tend to dominate when viewing all. Some days it gets to where looking at all isn’t very different than just looking at Memes@lemmy.ml or 196@lemmy.blahaj.zone.
Before someone says “you can just block communities you don’t want to see,” it’s not that I never want to see them, it’s that I want to be able to have a view that shows me what is new and popular in a wide variety of communities. I appreciate seeing a few good memes in my feed. The problem is when that’s all I see. Changing the sort from active to hot or top x days doesn’t have much effect on which communities dominate, so that isn’t the solution either.
“You can just subscribe to communities you like”. True, but that has the effect of narrowing what I see. I’d like a view that showed me new things I never thought to subscribe to.
Lemmy devs - if you are reading this - it would be nice to have a feed that limited the number of posts showing up from any particular community. It could be a simple cutoff of 2 or 3 posts, or maybe some sort of weighting function to cause additional posts from the same community to appear lower in the sort order for that feed.
I’d love to hear what devs and other users think about this.
Edit: To everyone saying “just sort be new” - yes, that has its uses, but it only solves part of the problem. I’d like a feed that shows me what is new and popular, but from more than just one or two communities.
If a community has grown a lot, unfortunately it takes up all the posts on the homepage. If I were a developer, I would sort by weighted success. For example, if the “x” community has 1000 subscribers and the post gets 100 likes, it has ten percent success. If the other ‘y’ community has 100 people and gets 11 likes, it has 11 percent success. This overrides the post in community x because the post in community y is more successful. This is the logic of ‘weighted success’. With this logic, a better ranking formula can be created.
deleted by creator
The thing is, a 10% upvoted post on a 10,000 people community is more popular than a 90% upvoted post on a 1000 people community - those 10,000 people in the former community are 10,000 people interested enough in that kind of thing to subscribe, whilst only 1000 people are interested enough in the other kind of thing.
So it does make sense to put the former higher up in the global page when sorted by popularity because globally that post was more popular.
However I do think there should be someway to as a user push down posts from certain communities without outright blocking the whole thing: maybe som throttling-down based on the rate of posts per time (i.e. the upvote threshold for posts from a community to come out in All depends on the number of otherwise qualifying posts in the last X days/hours, thus explicitly targetting the “flooding with posts” itself) rather than the straight count of upvotes or the proportion of upvotes that you suggest.
That said in the meanwhile I’m really tempted to block the more generic meme communities.
General popularity is not a good metric IMO. If I like a community, then it shouldn’t matter if a million people like or it’s only me and my cousin. If the community likes the content, I want to see it.
It’s trust between the members of a community.
However, weighted sorting is not a solution too, upvotes counts are not linear. Maybe, quantile sort?
What’s liked by the general population is a good metric for providing general stuff to the general population and that’s what we’re talking about in All.
That average can however deviate a lot from the sweet spot for some people, quite possibly a large minority (even the majority depending on how concentrated or not people’s tastes are around it).
Something that looks at your previous choices (or even generally stated choices in the form of communities you subscribe to or block) similarly to what some search engines and some social media sites will do, can shift that toward more your own specific tastes, but that’s computationally more expensive and requires more users and more user data to get better results (basically it’s finding certain kinds of users and local minima which are more satisfactory to them).
I suspect something like an AI solution (not LLM, just a much simpler neural network) running on your own device that tries to predict what you’re going to click on and learns with what you do (or not) is the only way for a personalized “no fluff on my feed” solution, but that’s for apps running on top of Lemmy, not the Lemmy engine.