mas.to is one of the many independent Mastodon servers you can use to participate in the fediverse.
Hello! mas.to is a fast, up-to-date and fun Mastodon server.

Administered by:

Server stats:

13K
active users

Alright, we appear to have uncovered something with the help of y'all.

Twitter had a featured apparently called "Quality Filter" - this was probably turned on at some point unbeknownst to me once I crossed X-number of followers but was opt-in otherwise.

This was undoubtedly the thing that hid nasty replies and prevented them from showing up in my notifs. And if this wasn't on for you (or you didn't have access) that may explain why my lived experience Over There has been easier than here.

@TechConnectify that’s a solvable problem then. In principle that could be implemented in mastodon.

@hyenagirl64 @TechConnectify but it really shouldn't be. Most of us are here to get away from the algorithm driven feeds, implementing "quality" filters would likely mean smaller accounts would start getting silenced, just because they don't have enough followers to be relevant to the algorithm. And if it's not automated, then you'll just get more drama, like entire servers being blocked because of a single admins opinions.

Technology Connections

@indirectferret @hyenagirl64 That's a bad read of what this did.

I saw stuff from small accounts *all the time* - it's was absolutely not as if you needed to have so many followers for me to notice your reply.

It was that when someone used abusive language, their reply would not get to my eyes.

Of course the filter will be imperfect and there will be collateral damage, but welcome to life.

@TechConnectify @hyenagirl64 don't get me wrong, I understand the desire to have some kind of filter. While I'm sure it's not to the extent you have to deal with, I get plenty of abusive messages sent my way as well (mostly just because I have a trans flag next to my name)

And I'm not totally opposed to there being some kind of user end filtering, I just worry that opt-out server side filters like twitter uses toe dangerously close to the line of having someone else decide what you can and can't see.

@indirectferret @TechConnectify @hyenagirl64
User-driven content filtering has always been a goal of the Mastodon project, from what I understand, as opposed to corporate driven algorithms. I'm not sure what a feature could do beyond censoring specific words and phrases, but perhaps this should be a feature request?

@RedstoneLP2 @Raccoon @indirectferret @hyenagirl64 That would probably be the easiest way to keep the spirit of Mastodon (entirely user-generated feedback) but it does have abuse problems. And the general HOA vibes here make me think abuse would absolutely happen under that paradigm.

But some sort of signals are necessary.

@indirectferret @TechConnectify @hyenagirl64

How would you think of solving this problem? There is the idea of free-speech, and then there is the idea of i-can-only-take-so-much.

@indirectferret @TechConnectify @hyenagirl64

I don't want someone else deciding what I can't see, but I would love to get help from people I respect deciding what I do and don't see.

Twitter had one algorithm, and it was all about maximizing engagement, and it ended up maximizing outrage. There are a million other possible algorithms, and I want people to be able to say "Try Janette's 'Fleeble' algorithm! I did, and I'm loving my feed!".

@TechConnectify You basically need a spam filter for notifications. I think this is a great idea. And should be relatively easy to implement. I mean spam filters already work quite well on emails. You "just" need to train it with a few dozen messages each of what you DO and DON'T want to be notified about. Question is: Who will implement this for you?

@root42 @TechConnectify Replies and notifications, yeah. Theoretically could be done client-side, like Thunderbird does, so if someone wants to fork the official app and add it, they could do so without any changes to Mastodon (the service) itself. Everybody could train their own classifier, or download someone else's if the classifiers are sharable.

Honestly this seems like a huge quality of life improvement