this post was submitted on 06 Jul 2023
922 points (95.4% liked)
Fediverse
28724 readers
113 users here now
A community to talk about the Fediverse and all it's related services using ActivityPub (Mastodon, Lemmy, KBin, etc).
If you wanted to get help with moderating your own community then head over to [email protected]!
Rules
- Posts must be on topic.
- Be respectful of others.
- Cite the sources used for graphs and other statistics.
- Follow the general Lemmy.world rules.
Learn more at these websites: Join The Fediverse Wiki, Fediverse.info, Wikipedia Page, The Federation Info (Stats), FediDB (Stats), Sub Rehab (Reddit Migration), Search Lemmy
founded 2 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
If the βhotβ and βactiveβ filters continue to work as expected and bots get reasonably moderated or blocked, I donβt even think Lemmy needs a high barrier to entry or petulance. The most important thing is to not optimize any recommendation or sorting algorithm on session duration, ads seen before closing session, and revenue per user.
This is so true and people seem to have a really hard time seeing this. The cultures on other social sites are far more manufactured than we'd like to believe. I think the human driven systems of Lemmy and Mastodon are brilliant but the true killer feature of the fediverse is going to be an open content recommendation algorithm. A collectively developed non-profit driven algorithm would undoubtedly be better at surfacing positive impact content than either system.
Can we use ai to judge emotional content of threads so I can get recommendations for threads where people are relaxed and happy?
Actually yes and I donβt think itβs a bad idea. Sentiment analysis is not a hard task nowadays. If overused the site would become a bunch of artificial positivity, but I think there could be a place for this. Could be part of a mod toolkit too.
I definitely don't think something like that should be used to only show one emotion though. If AI were used to control content based on how it makes people feel it should try to balance, not control how we feel. Give us an equal amount of everything.
It's not good to cloak ourselves in only feelgood stories and lies that sound nice and ignore all the bad stuff just because an algorithm wants us to feel nice and cozy. If no one cares about the bad stuff, the bad stuff gets worse.
I'd imagine if it uses the sentiment as a weight but not a gate when selecting what content to boost that would probably provide a good balance. Especially if you get some controls to adjust the weights in the settings panel
Unironically, yes absolutely.
Well then, I unironically want that!
Too much arguing in discussion forums and not enough relaxed fun.
Completely agree that this is where the really exciting potential is, but equally a potential for misuse as algo development will be a black box to most.
Funny enough Hot filtering is currently bugged on small instances.
Yesterday it was bugged on world, too. I'd keep hitting next page and seeing half of the same things on every page. It's working better today though.
It's been noted on the GitHub. It seems the bug is amplified on small instances. I'll regularly get posts from two weeks ago with no comments on the top of All/Hot when using this account.