If the “hot” and “active” filters continue to work as expected and bots get reasonably moderated or blocked, I don’t 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.
Actually yes and I don’t think it’s 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.
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.
If the “hot” and “active” filters continue to work as expected and bots get reasonably moderated or blocked, I don’t 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 don’t think it’s 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.
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.