Something I’ve thought about a bunch re: recommendation engines is the idea of a “sweet spot” that balances exploration and safety

Though actually I should start by saying that recommendation engines tend to aim to maximise engagement, which is why manosphere type content is so prevalent on places like YouTube if you go in with a fresh account — outrage generates engagement far more reliably than other content. I’m imagining a world where recommendation algorithms may be able to be individually tailored and trained, where I can let my goals shape the recommendations. I did some tinkering with a concept like this in the context of a personal music recommender, and I gave it an “exploration” slider, where at maximum, it’d suggest some really out-there stuff, but lower down might give me new songs from familiar artists. That project worked quite well, but it needs a lot of work to untangle before I can figure out how and why it worked so well.

That was a super individualistic program I made there, in that it was trained exclusively from data I gave it. One can get individual goals without having to rely on the data of just one person though - listenbrainz is very cool — its open source, and they are working on recommendation stuff (I’ve used listenbrainz as a user, but not yet as a contributor/developer)

Anyway, that exploration slider I mentioned is an aspect of the “sweet spot” I mentioned at the start. If we imagine a “benevolent” (aligned with the goals of its user) recommendation engine, and say that the goal you’re after is you want to listen to more diverse music. For a random set of songs that are new to you, we could estimate how close they are to your current taste (getting this stuff into matrices is a big chunk of the work, ime). But maybe one of the songs is 10 arbitrary units away from the boundary of your “musical comfort zone”. Maybe 10 units is too much too soon, too far away from your comfort zone. But maybe the song that’s only 1 unit away is too similar to what you like already and doesn’t feel stimulating and exciting in the way you expect the algorithm to feel. So maybe we could try what we think is a 4 or 5. Something novel enough to be exciting, but still feels safe.

Research has shown that recommendation algorithms can change affect our beliefs and our tastes [citation needed]. I got onto the music thing because I was thinking about the power in a recommendation algorithm, which is currently mostly used on keeping us consuming content like good cash cows. It’s reasonable that so many people have developed an aversion to algorithmic recommendations, but I wish I could have a dash of algorithmic exploration, but with me in control (but not quite so in control as what you describe in your options 3). As someone who is decently well versed in machine learning (by scientist standards — I have never worked properly in software development or ML), I think it’s definitely possible.

I personally wouldn’t mind algorithmic recommendations if:

• you can control or choose the algorithm
• you can turn it off, or it turns off after you follow N amount of users

Discovery is important when you’re initially signing up, but once you found the people you want to follow, you don’t really need it any more. It should just be there to help new users, essentially. As long as it’s open source and not run for profit, there’s not the traditional incentive to keep your eyeballs on the app like we see with the other networks.

On art sites - i.e. DeviantArt and the various furry sites that copied it in the 2010s (many of which are still very much active) - every user has a “favorites” page to which they can add anyone else’s posts. This also sends a notification to the artist. When I get a favorite or a comment, I always check that artist’s posts and their favorites page, to see what else they have favorited, and I can reliably find new people and posts that way. I don’t think this is really analogous to a boost, since it doesn’t show up in your feed, but it’s also not quite the same as “likes”, since it’s much more expected that people will go looking through it (people don’t really look at other people’s likes much, even though they are almost always public). Plus it’s almost guaranteed that it’ll be filled with art, as opposed to blog or microblog style text posts, so you’re not as subjected to the hot takes of random people you don’t know.

Given the way things are in my perspective, what I want on mbin & lemmy is somewhere like a mix of 1 & 2, with 3 as a solid option. I know that the torches and pitchforks are about to come out, but I’ll try to outline the way I see it.

When I’m in a meme-scrolling mood, I have to look up meme magazines / communities to start (Method #3). Fine, that’s working as intended. Obviously that will lead naturally to Method #1; as I subscribe and gradually follow other posters, my bubble will grow.

But what I want for the ‘threadiverse’ is a more unified suggested page. If I’m in, let’s say, [email protected], I’d like to also have my feed show content from [email protected], or lemm.ee, or whatever other threadiverse instances that my chosen instance is federating with. I’d also like to see “subject memes” on my meme feed as a default - Science Memes, Star Trek Memes, etc… That falls under Method #2 - because I want the software to predict that because I’ve subscribed to memes@*, and interacted with content from memes@*+1, that I will also like *memes*@*. Obviously this could also be a matter of tagging and magazine integration, but that’s something that would help the fediverse feel more united and less daunting for people.

Obviously dealing with the microblog side, mandatory tags or some form of community selection would be great to help out. It would be nice to see more microblog entries from Mastodon, Misskey, Pleroma, etc., sorted into magazine-like collections by tags.

Looking back at my own life, I found the first few online communities I ever seriously joined (when I was a preteen, for context) through a web search, then discovered most others (recursively) from there, until I ended up (among other places) here on lemmy (which I can trace back to reddit, which I can trace back to a forum I started to pay attention to because of one of these original online communities preteen me found through a web search; not providing more info for privacy reasons). :P

So #1 and #3 are how it should work, IMHO, although #3 mostly for people who aren’t yet engaging with anything at all, most things will be discovered through #1.

I think most people use the Internet not for posting anything (or at least not much) themselves, but for looking up things they want to know (through a web search). In the pre-smartphone era, web searches would often direct to specific websites which might have forums attached to them, that was how I first started to seriously engage in my first online community actually. This isn’t the case much nowadays: many search results are either wikis (which are communities themselves, but don’t really invite discussion that isn’t about working on the wiki) or blogs/WordPress websites which may or may not have a comment section, but it’s relatively rare for them to have forums or even to link to reddit/fediverse communities to discuss their subject matter.

So I think it would be desirable if we managed to change that last part: top search results for many terms on search engines should be, or link to, fediverse communities, which should make it clear that users are invited to join. That would help us get more users engaged with fediverse communities in the first place, they would naturally discover more communities once they’re here.

Search: you specifically attempt to find what you’re looking for through some search capability. Big tech is pushing against this more and more.

Funny, considering that you cannot find threads, communities, etc. that aren’t already federated with your instance because of someone being subscribed to it beforehand.

I wish we had a nice tagging system (and I don’t think they should be hashtags) that was also in common use.

I want to be able to search any post related a certain topic, and sometimes, these may not always be in that topic’s community, because topics can overlap. For example, I might want to read posts about Ukraine war, but those might be in world news, US news, or combat footage communities. Could be a community about Ukraine in general, or Ukraine war specifically.

I also may not want to get it from a single Ukraine community. Maybe by finding posts with the “Ukraine war” tag, I’ll see several communities and join the one I want. But there needs to be a way to group them somehow.

Such a tag system may be useful for combined topics. For example, I may want to look for posts about music software. They might not be common in the music community, or software communities. But I could filter by both tags and find what I want.

For me, it’s full text search.

I tend to want to find an opinion on something very specific, so if I can just toss a phrase or model number or name of something into a search field and get actual non-AI, non-advertisement, non-stupid-shit results, that’d be absolutely ideal.

Like, say, how Google worked 15 years ago.