We use manual approval for programming.dev accounts where there is a very simple instruction you must follow to be approved. The amount of spam that fails that test makes me concerned about the amount of bots from instances without any barriers for account creation.
What happens on reddit (in regards to spam) will inevitably finds its way to ActivityPub link aggregators like lemmy.
I am sad that the current generation of federated social media/networks still doesn’t have much, if any, implementation of web of trust functionality. I believe that’s the only solution to bots/AI/etc content in the future. Show me content from people/accounts/profiles I trust, and accounts they trust, etc. When I see spam or scams or other misbehavior, show me the trust chain connecting me to it so I can sever it at the appropriate level instead of having to block individual accounts. (e.g. “sorry mom, you’ve trusted too many political frauds, I’m going to stop trusting people you trust”)
I would definitively use it if it was implemented. Make it work like it is in GPG, where you can rank users based on your trust, and that is then propagated to others.
This concept reminds me of a certain browser extension that marks trans allies and transphobic accounts/websites using a user aggregate with thresholds that mark transphobes as red and trans allies as green.
I guess the question is how specifically you implement such a system, in this case for software like Lemmy. Should instances have a trust level with each other? Should you set a trust when you subscribe to a community? I’m not sure how you can make a solution that will be simple for users to use (and it needs to be simple for users, we can’t only have tech people on Lemmy).
For the simplest users, my initial idea is just a binary “do you trust them?” for each person (aka “friends”) and non-person (aka “follow”), and maybe one global binary of “do you trust who they trust?” that defaults to yes. anything more complex than that can be optional.
You’d see posts in a community/group/etc based on your trust of the community, unless you’ve explicitly de-trusted the poster or you trust someone who de-trusts them (and you haven’t broken that chain).
Right, so if I have no connection to someone else, it’d be “neutral” and I’d see the post. If I trust them transitively, then it would be a trusted post and if I distrust them transitively, it would be a distrusted post.
I think implementing such a thing would not only be complicated but also quite computationally demanding - I mean you’d need to calculate all of this for every single user?
Yes! Web of trust is the only way. Everything else can be scammed. I am kinda wondering if it could be invites and if severing could be automated for social media. “We just banned a third person who came in on your invitations. Goodbye.”
Honestly I already believe that this has happened.
My reason for thinking this is because of this:
The spike that happened on October 2023 after the initial spike that happened due to the Reddit protests seems unnatural to me.
Someone gave the explanation of the release of the mobile clients but even then I wouldn’t think it would lead to a spike equivalent to the initial one since it would mostly just be people using an account they already had instead of creating a new one.
Like honestly if someone knows what event happened then that made so many new users join I’d appreciate it.
I didn’t get into Lemmy until there was a mobile client available, Sync to be specific. I believe it since a lot of Reddit users were basically mobile only. So, for a few months I basically subsisted on YouTube alone.
I feel your pain. I also tried Instagram to satiate the mobile scrolling, but the comments there are just horrible and low-effort. The fediverse via Sync is okay, but there’s still much to be desired.
Sync is only one of many clients. And when it came out wasn’t the point of my post. But if no big clients came out in October then they couldn’t have been a factor at all.
We use manual approval for programming.dev accounts where there is a very simple instruction you must follow to be approved. The amount of spam that fails that test makes me concerned about the amount of bots from instances without any barriers for account creation.
What happens on reddit (in regards to spam) will inevitably finds its way to ActivityPub link aggregators like lemmy.
I am sad that the current generation of federated social media/networks still doesn’t have much, if any, implementation of web of trust functionality. I believe that’s the only solution to bots/AI/etc content in the future. Show me content from people/accounts/profiles I trust, and accounts they trust, etc. When I see spam or scams or other misbehavior, show me the trust chain connecting me to it so I can sever it at the appropriate level instead of having to block individual accounts. (e.g. “sorry mom, you’ve trusted too many political frauds, I’m going to stop trusting people you trust”)
I think this would be a great feature request: https://github.com/LemmyNet/lemmy/issues
I would definitively use it if it was implemented. Make it work like it is in GPG, where you can rank users based on your trust, and that is then propagated to others.
This concept reminds me of a certain browser extension that marks trans allies and transphobic accounts/websites using a user aggregate with thresholds that mark transphobes as red and trans allies as green.
I guess the question is how specifically you implement such a system, in this case for software like Lemmy. Should instances have a trust level with each other? Should you set a trust when you subscribe to a community? I’m not sure how you can make a solution that will be simple for users to use (and it needs to be simple for users, we can’t only have tech people on Lemmy).
I think that there would probably have to be some database, somewhere, to store that data?
For the simplest users, my initial idea is just a binary “do you trust them?” for each person (aka “friends”) and non-person (aka “follow”), and maybe one global binary of “do you trust who they trust?” that defaults to yes. anything more complex than that can be optional.
But how does this work when you follow communities? Do you need to trust every single poster in a community?
You’d see posts in a community/group/etc based on your trust of the community, unless you’ve explicitly de-trusted the poster or you trust someone who de-trusts them (and you haven’t broken that chain).
Right, so if I have no connection to someone else, it’d be “neutral” and I’d see the post. If I trust them transitively, then it would be a trusted post and if I distrust them transitively, it would be a distrusted post.
I think implementing such a thing would not only be complicated but also quite computationally demanding - I mean you’d need to calculate all of this for every single user?
Yes! Web of trust is the only way. Everything else can be scammed. I am kinda wondering if it could be invites and if severing could be automated for social media. “We just banned a third person who came in on your invitations. Goodbye.”
Definitely something that will emerge in the future once we’ll inevitable get bots here too
That sounds smart!
That would be awesome, cause you would still be able to see your mom’s post!
Honestly I already believe that this has happened.
My reason for thinking this is because of this:
The spike that happened on October 2023 after the initial spike that happened due to the Reddit protests seems unnatural to me.
Someone gave the explanation of the release of the mobile clients but even then I wouldn’t think it would lead to a spike equivalent to the initial one since it would mostly just be people using an account they already had instead of creating a new one.
Like honestly if someone knows what event happened then that made so many new users join I’d appreciate it.
I didn’t get into Lemmy until there was a mobile client available, Sync to be specific. I believe it since a lot of Reddit users were basically mobile only. So, for a few months I basically subsisted on YouTube alone.
I feel your pain. I also tried Instagram to satiate the mobile scrolling, but the comments there are just horrible and low-effort. The fediverse via Sync is okay, but there’s still much to be desired.
Is your issue with Sync or with the current state of the fediverse?
The current state of the fediverse. Sync is great!
Sync came out in August tho, I used Connect in the meantime lol
Sync is only one of many clients. And when it came out wasn’t the point of my post. But if no big clients came out in October then they couldn’t have been a factor at all.
That is sorta my point, I don’t know how much mobile app releases factored in to the spike back up in October. It probably did have some effect though
Newer user here… the api stuff got me to delete my reddit account but still surf it, it was the day of the IPO that i created my lemmy account…
Welcome! Let us know if you have any question.
Welcome new person, best start brushing up on beans and jeans if you want to survive.
That happened in March 2024 I think. And Reddit filed for the IPO in December 2023.
https://www.reuters.com/markets/deals/reddit-seeks-launch-ipo-march-sources-2024-01-18/
The day it IPOd on the market was my final day not the day of filing… I was holding out hope it wouldn’t happen lol…
Fair enough
Is that just accounts in total or active accounts?
I didn’t comment much in the beginning.
Now I try to comment at least once a day.
accounts in total.
Wait, then how would it go down? Are people deleting their accounts that much?
Entire instances disappear as well.
Apparently. But it seems like it only happened around the beginning after the second spike it stabilized for some reason.
Edit: Here is the page with the stats
Okay then I will admit, it does seem fishy.