A disappointing part about #Mastodon is when someone #boosts / #likes you, and you visit their #profile and you discover you share a lot of similar interests!
So you #follow them - but they don't #followback.. and you never hear from them again..
..so you get in your own head...
Dont let this get you down - the #fediverse is new for a lot of us!
Like.. follow.. unfollow .. whatever! Just interact in good faith with no malice
Blame the #UX for now and know it will improve with time! 😁
@multiverseofbadness The lack of prioritization features in most #ActivityPub implementations makes following lots of people somewhat obnoxious, as it ends-up effectively as practical to read everything as it is from the main federated timeline.
So I end-up following very few people and simply checking up on my followers manually from time to time.
It's very much a UI-specific problem and could certainly be addressed, but I've yet to do any such thing.
@lispi314 @multiverseofbadness
It's almost like those algorithms actually did provide value 🙂
I'm half joking, because for everyone around here celebrating that there's no algorithm here other than chronological, the experience really could be improved by having some other options for users.
@lispi314 @multiverseofbadness
Well to be serious, I think the issue is people who lump all algorithms together except for raw chronological, ignoring for one that chronological is itself and algorithm, and also overlooking that there are some algorithms that are better and some that are worse at meeting any particular user's wants.
I'm just laughing because I see so many people on this platform flat out rejecting the idea of allowing any algorithms at all, like luddites rejecting the idea of any technology at all, seeing it all as negative.
Sounds like we probably agree that user choice in algorithms would probably be a positive thing, even though a lot of people would be really upset about going in that direction.
@volkris @multiverseofbadness I think we do agree. I'd keep any such algorithms strictly under control of the user in their user-agent, whether that be the WebUI or some alternative client.
The backend has no need to trouble itself with that (and indeed, not doing so would avoid additional load).
First, I agree that the word "algorithm" is overused and often misused to mean things that it doesn't. Unfortunately, evangelists and journalists don't make an effort to explain this.
Implementing good algorithms beside chronological is hard. I have build my own with bovine. My decision was not to bother making a smart algorithm instead make content easy to dismiss, i.e. swipe left on phone.
Next: The questionable thing would be to train a neural network to select the timeline. This runs into all kinds of issues, like copyright, and what metric to optimize for. User happiness is not something you can measure. Instead, something like "number of interactions" is used, which results in a hellscape of confrontation for obvious reasons.
@helge @multiverseofbadness @volkris I think the misuse is from a lack of adequate education in maths. "Sorting algorithms are bad" is obvious nonsense. Specific ones /are/ bad and ones based on statistical heuristics without user-control also are. Not because of heuristics, but because of the lack of user control.
Regarding interactions & confrontation hellscapes, allowing users to weigh interactions negatively so as to make them less likely would help mitigate that.
@lispi314 @helge @multiverseofbadness
Personally, I wish the heuristics would be built on a cryptographic Web of Trust framework to bring external social sorting and identity verification into the mix, but that's just something I yell into the void occasionally :)
@volkris @helge @multiverseofbadness That would rely on people intentionally opting into signing their messages or otherwise authenticating them.
External social sort also implies that others' preferences would override mine (whether directly or through added opportunity cost) which I wouldn't want.
@lispi314 @helge @multiverseofbadness
Meh, the signing would simply be built into the account. It could even be just a signed certificate associated with the account and not each message if devs didn't want to deal with that extra step.
The only opt-in would be people vouching for each other.
But to clarify, by external social sort I mean that profiles would end up with notations saying "your friend says this is a real person who's cool to talk to" vs "you've no connections to this account so it might be a spammer or troll"
You'd still have every ability to do with that what you want, whether you want to ignore the friend-of-a-friend connection or take it into account.
@volkris @helge @multiverseofbadness Ah, I had been thinking the intent of the cryptographic WoT was to make reputation of a given individual portable across instances, rather those aspects pertaining to more baseline federation.
Hence that bit about verification.
That implementation of yours does sound more useful and benign.
@volkris @multiverseofbadness The algorithms' main flaw is quite simply that the user isn't in control and cannot change them. Something which respects the user like scoring (https://www.gnu.org/software/emacs/manual/html_node/gnus/Scoring.html) would be quite fine.