RE: https://mastodon.social/@hfalcke/117455063326697336
Yep. Me 3 years ago:
https://hachyderm.io/@mekkaokereke/110791346726665161
"Who controls the algorithm, and what is it optimizing for?" is a more useful question than "chronological or not?"
RE: https://mastodon.social/@hfalcke/117455063326697336
Yep. Me 3 years ago:
https://hachyderm.io/@mekkaokereke/110791346726665161
"Who controls the algorithm, and what is it optimizing for?" is a more useful question than "chronological or not?"
@mekkaokereke I suggested multiple open, tunable algorithms on here like two years ago and got dragged for days. Fedi's blind spots can be really tiresome.
Yeah, my impression is that a lot of non-technical users have picked up the word "algorithm", but take it to mean something like "corporate voodoo", and so they wind up having a quite different discussion than anyone with an informatics background.
I did once manage to have a discussion about what the word means with someone, but I was really feeling Brandolini's law the entire time.
@syklemil @MichaelTBacon @mekkaokereke Yeah, algorithm as black box rather than just "hey, here's what we use to prioritize"
I think that Lists would be a good place for that.
I enjoy mentally taking features to their logical extreme and working backwards from there and the logical extreme for this I think would be a text field where you can put in a Lua (or similar language) script that orders the posts that match the list's filter. Then people can share their algorithms and tune them together.
@MichaelTBacon @mekkaokereke how about no algorithm, since none is necessary and there is no such thing as a useful algorithm because it's technically impossible for a machine to read my mind and know what I want to see in that very moment. Only techbros want illogical bullshit like this to wank to. But the techbro sites are x and bluesky.
Respectfully, this is a misunderstanding of what an algorithm is. I would invite you to see some of the thoughtful replies to my original post.
@MichaelTBacon @mekkaokereke this is also solvable with just better client views, like Catch-Up when using Phanpy.social -- still chronologically sorted but the volume-difference managed
@mekkaokereke Strongly agree, we need more open discussion about non-chronological feeds that are open and beneficial. This can be easily be done and have a positive impact on the community.
@scottjenson @mekkaokereke Alternatives need to not only be open, but also be chosen by the user without penalty like network changes.
@mekkaokereke That's one reason I think federated algorithms aren't a terribly good solution: The fact of externalizing timeline control makes it easier to obscure who controls an algorithm and what it's optimizing for. It's one of the reasons Bluesky doesn't appeal to me.
From my POV, a better solution would be a frontend "building block" system for constructing your own algorithm, maybe with an import option for sharing patterns in a human-readable format. That way, you could check an algorithm before applying it to your account, and no one could alter it remotely or without your knowledge.
@lrhodes @mekkaokereke I would love to have this feature! I do make use lists for less frequent posters and filters and I turn off boosts for people I follow who boost so much it takes over my whole feed but more control is always welcome.
@mekkaokereke I think whatever approach ends up happening it should be made abundantly clear that no matter what happens it won't be built around engagement so if a non-chronological feed becomes an option it isn't going to trigger the kinds of shenanigans we see in the other socials.
Behind all of it is the unquestioned assumption that i want or need to read "everything".
I miss all conversation that occurred before i walked into the room. This is how the world works. So what?
Tech heads wanna tech.
@mekkaokereke me too around that same time I think! I feel heard now 😅
I think the key is algorithmic flexibility. Have a few simple and sane defaults , and give the user a way to deeply tune their preferences (of course the nerd in me would ask for a python interface of some kind, but I get this might not be mainstream enough!)
@mekkaokereke I would build a feed that attenuates both ends of the spectrum.
@mekkaokereke I think the answer is somewhere in between. Who controls the algorithm is relevant, but offering sane defaults is also important.
I'm not opposed to users choosing/tuning the algorithm for their feed, but this is not trivial and it's important to have save options. Addictive choices are addictive. That doesn't change, just because users can choose this themselves.