cross-posted from: https://lemmy.world/post/4150119
The assumption is that centrally managed social media is bad because their algorithm is bad. But actually, they are bad because they are centrally managed and force one algorithm onto you. I’m not even advocating algorithm-by-choice. Even instance-specific algorithms would already work and would make the whole experience much more enjoyable and less boring. And if an instance’s algorithm(s) is too aggressive, it gets defederated. That would result in a much more exciting experience imo. And by the way: what’s the problem with getting old posts back in the timeline if it makes the overall conversation more interesting?
The assumption that an algorithm is bad is prevalent here, although that is not the only reason why centrally managed social media is thought of as bad.
That being said, algorithms I think would be very beneficial for this platform if made correctly, with psychology in mind. Algorithms on other platforms serve to addict users primarily, and as a byproduct reinforce confirmation bias. We here don’t have an algorithm so no addiction, but there are still downsides. Content like lemmygrad or hexbear put out is shown to those who really don’t want to see it, and the only option is to completely block them which is permanent, and doesn’t allow change over time (assuming this platform is built to last decades this will happen). additionally the vast majority of lemmy is liberal or whatever the corresponding ideas are seen as in the eu. This is dangerous as it reinforces our confirmation bias as well, allowing the platform to devolve into extremity over prolonged periods of time. This can already be seen happening by what some people comment and get support for under political posts regarding United States current events, and other political debates. I would provide evidence for this but I’m on my phone and can’t go find some without cancelling out of this post, if you ask I will find some. The extremity that will be caused by this while it isn’t good for us, may not necessarily impact the subs you subscribe to or you at all, this is good for you but it still damages the platform. If we wish to grow we need to be more welcoming to new users, this means more than just the sign up. If new users see constant extremist ideas under political posts they see in all and may want to engage in, or extremist ideas under memes in subs like fuckcars or atheistmemes they will feel less like a member of the community and more like an outsider and will leave. These issues are not our fault as they are just how the community has evolved and basic humanity at play, but they are still issues. An algorithm could solve these. If we design an algorithm to show users content they are interested in and content that they may engage in more often new users will feel connected quicker, however we still have the current and algorithmic confirmation bias issues. These can be solved by recommending more and more stuff that they don’t agree with or aren’t as interested in and boosting this content for them specifically so that it shows up less than the stuff they are interested in but is still present. This can be with post and comments. The obvious issue with this is they won’t interact with it if they don’t like it, however if done right you would boost content based on what is explicitly says and not what it may imply after deep thought, so all they have to do is see it and their confirmation bias doesn’t take effect as much, and they won’t get addicted as they won’t see only stuff they are interested in. The other issue with this is it requires collecting data. Since this would be a significant change to lemmy as code, I would propose that this data be collected locally and encrypted before being uploaded to your instance. This way the data would be secure but also accessible to other front ends that you log in through, because data collection isn’t inherently a bad thing if it serves to solve issues caused by our brains inherent imperfect compatibility with social medias and the internet. All in all I think this could be a very good thing that would not only make browsing lemmy more interesting, but would make it fight against the flaws in our brains.
What do yall think about this idea?
I like some ideas, other I don’t.
These issues are not our fault as they are just how the community has evolved and basic humanity at play, but they are still issues. An algorithm could solve these.
For me, this sounds like solutinism. Its a societal problem/problem by the network, which (imo) needs to be solved by the network primarily (if it is even a problem). You just want to solve it with an algorithm that evens out political differences and the flaws of our brain? The question is who is implementing this algorithm then. This would be a question for any algorithm and its a debate that should defintely included in the debate around algorithms but I think overuse is dangerous as well as giving too much responsibility to the algorithm where it doesn’t necessarily belong.
If say, christians want to start a sub-feed here and are continously attacked and made fun of, that’s primarily a problem of the network and the moderators and the moderation tools of that specific instance and shouldn’t be tried to be solved with an algorithm.
However, if we are speaking about implementing algorithms in the fediverse, it would definitely worth to consider whether they help some political views or what their effects on the network are in general. And I think it good be a valid point that while chronological order is easy to implement, to understand and fair, it could have some negative networking effects as well like you described.
So basically, I think its an interesting point and I would even go as far as agreeing with you with the remark that it would be kept an eye on not putting too much faith in the algorithm and giving the problem at the end of the day to the community. The algorithm should be politically neutral, encourage the discovery of new content and be welcoming to new users. I agree that especially the second one is a central problem of the fediverse, the third one kind of, and the first one will need to be kept when trying to achieve the others. A balance between these factors would be for me the ideal algorithm, but if you give the possibilitiy of implementing algorithms to the admins (which I would think would be good) it would be a matter of the communities discussion and the overall fediverse, which I think would only be sensible.
Maybe this is also something that the Fediverse just isn’t ready for yet. But it will definitely become a more pressing issue at some point in the future. If the Fediverse wants to be a place for relevant debate, the chronical feed will need to be replaced with something better.
The algorithm should also somehow prioritize posts for you to encourage interaction, because else as you pointed out, the will reduce your amount of acounts you follow, which shouldn’t be the case AND else you also have a dead timeline, where no post has any comments, which I have often enough on Mastodon
On the site formerly (and still) known as Twitter, I missed a lot because it did not show me everything from accounts I followed.
On Mastodon, I miss a lot because I wasn’t browsing my feed at the right time.
This would matter less on Mastodon if my feed was endless and I could browse it until I’d caught up with everything new. But it is not endless, so I can’t.
I don’t see how you can have multiple instances all injecting stuff into my feed, which is what the (other) OP seems to be suggesting? I don’t want any kind of algorithm deciding what I see and definitely not some kind of hyper-competitive inter-instance competition for my attention. But I wouldn’t mind if Mastodon worked out what I hadn’t yet seen from the accounts I follow and put it into my feed while I’m actually browsing it.
