I think a big problem behind the reluctance of alternatives to the strictly chronical timeline on Mastodon is that people fear that too much power is taken out of their hands if they are introduced. But the fact is: it is already the case that we put a lot of trust in administrators to put the correct software in place. A strictly chronological timeline makes one thing less to worry about but basically, it only reduces the symptom of the problem. Instead, I think the real problem needs to be faced: take away the fear of users that their instances are not working as they are supposed to and give them the power to check themselves whether the instances they are on are actually doing what they subscribed to.

As the most important, I think of the following two:

Defederation Tool: shows from which other instances your own instance defederated (I think that already exists).
Timeline tool: is the timeline curated based on the algorithm the instance proposed.

If these are in place, you could check that you see the right posts by the right instances, which is already a nice thing to know to begin with and would at least me quite content for introducing custom timelines and thereby giving more power to the admins. And with the mentality of this being an important issue, there would always be someone trying to see if an instance is run as promised and most admins wouldn't bother trying to do bad things.

Additionally, the algorithms would need to be determinstic and data collected by the instance about the user downloadable.

PS: Of course admins are doing a great job here, also given that most of them are volunteers. I'm not saying they are bad people, I'm just saying there need to be tools to control what they are doing if more powerful tools will be introduced to them in the future like custom timelines.

all 25 comments

sorted by: hot top controversial new old

[–] waffle@sh.itjust.works 5 points 1 year ago* (last edited 1 year ago) (1 children)

No. If you want to run an algorithm you should run it on your computer and not on somebody else's.

Ignoring the fact that having a proper recommendation AI for every single user would be environmentally disastrous, it would also place much more burden on the ones hosting instances. Keep in mind that most instances are hosted by people who do not earn anything from them and that many bigger instances already had to rent bigger servers because of the influx of people. Adding computationally expansive algorithms in the mix would just increase the cost for the volunteers on top of signing the death of some smaller instances run on a tight budget.

It would also be prone to recreating the SEO mess that we can see today on social medias like youtube where, if you want to grow your community, satisfying the algorithm becomes more important than the actual content of your posts.

However, I would have no issues with an algorithm that a user of an instance could run on their devices and tweak to their liking. This solution would probably be less convenient but would avoid most of the mess.

EDIT: Sorry if my comment came out as too aggressive and thank you for making this post. I think that's an important issue to discuss and, as thanks for bringing it up, you have my upvote :)

[–] blue_berry@lemmy.world 4 points 1 year ago (1 children)

Ignoring the fact that having a proper recommendation AI for every single user would be environmentally disastrous, it would also place much more burden on the ones hosting instances. Keep in mind that most instances are hosted by people who do not earn anything from them and that many bigger instances already had to rent bigger servers because of the influx of people. Adding computationally expansive algorithms in the mix would just increase the cost for the volunteers on top of signing the death of some smaller instances run on a tight budget.

Ok, I didn't think of that :(

It would also be prone to recreating the SEO mess that we can see today on social medias like youtube where, if you want to grow your community, satisfying the algorithm becomes more important than the actual content of your posts.

I think federation should solve that right? If an instance has to aggressive algorithms it gets defederated and people leave.

However, I would have no issues with an algorithm that a user of an instance could run on their devices and tweak to their liking. This solution would probably be less convenient but would avoid most of the mess.

That would be cool

[–] waffle@sh.itjust.works 1 points 1 year ago* (last edited 1 year ago)

It would also be prone to recreating the SEO mess that we can see today

I think federation should solve that though?

Ah, yeah you're probably right my bad. Big data isn't my speciality so I can't say much about that :/

[–] pewgar_seemsimandroid@lemmy.blahaj.zone 3 points 1 year ago (1 children)

sure i guess lemmy made me realize i don't need an algorithm just a subscribed page

[–] blue_berry@lemmy.world 1 points 1 year ago

They aren't strictly chronological thought. Some basic filtering/sorting algorithm is what makes Lemmy much better in comparison to Mastodon imo

[–] Bristlerock@kbin.social 3 points 1 year ago* (last edited 1 year ago)

I have zero problem with curated or algorithmic timelines. I have a 100% problem when there isn't a chronology timeline option.

It's simple really: give me the permanent option of chronological without the dark pattern fuckery of having to reset it periodically, or fuck off forever.

[–] Anafroj@sh.itjust.works 2 points 1 year ago (1 children)

I would one up that suggestion : what could be awesome would be to allow users to choose which sorting algorithm they use, and possibly tweak it. This would allow people to share the sorting logic they like, and there would be no trust issue, since you can verify the logic is respected by changing it.

Not sure how realistic this proposal is, though, because this could lead to performance issues if users submit too complicated sorting logic, which could be exploited to DoS an instance. On the other hand, it could be solved with a timeout mechanism : "if your query takes more than 100 ms to load, we kill it". And also, you can't just let users run arbitrary SQL, obviously, so this would require to implement some sort of meta-language safely transpiled, this would be the real challenge.

[–] blue_berry@lemmy.world 2 points 1 year ago (2 children)

I would one up that suggestion : what could be awesome would be to allow users to choose which sorting algorithm they use, and possibly tweak it. This would allow people to share the sorting logic they like, and there would be no trust issue, since you can verify the logic is respected by changing it.

Bluesky already implemented that. I think its a cool concept although most users will just stick to the default one

[–] wagesj45@kbin.social 4 points 1 year ago (1 children)

Which would make it important that the default would be something with no tweaks, like a chronological list. But I would be all for a scripting setup or some other configurable sorting engine to play with on a per-account basis. Maybe you could even subscribe to other's sorting configuration.

[–] blue_berry@lemmy.world 0 points 1 year ago* (last edited 1 year ago) (1 children)

Which would make it important that the default would be something with no tweaks

No necessarily. Just let it federate.

Maybe you could even subscribe to other’s sorting configuration.

You mean subscribing to algorithms of other instances? Really interesting idea but I doubt that it would be a good idea security-wise.

[–] wagesj45@kbin.social 2 points 1 year ago (1 children)

It would have to be a defined standard. I imagine it not so much as a script that runs, but rather a description of fields and weights. I guess there might be some computation involved, but I think a standard could be devised that there would minimize the security risk.

[–] blue_berry@lemmy.world 2 points 1 year ago (1 children)

Interesting idea. Who would be setting the standards? Some neutral committee? W3C?

[–] wagesj45@kbin.social 1 points 1 year ago

Don't know. As long as it is open source, I'm not sure it would matter. Eventually one would win out with the community.

[–] Anafroj@sh.itjust.works 2 points 1 year ago (1 children)

Bluesky already implemented that

Oh really? Awesome. I'm still not going to use a generalist social network, but they're doing it right. :)

although most users will just stick to the default one

Yeah, indeed. My taste for tinkering may be showing, I should acknowledge it's not something widespread. 😅

[–] blue_berry@lemmy.world 1 points 1 year ago

I'm the same but the general user doesn't want to worry about this kind of stuff. They want to sign up to an instance, probably based on a recommendation by a friend, and have a good time. And if they don't, they change to another one. The algorithm would just blur into the whole experience of an instance.

[–] CaptObvious@literature.cafe 2 points 1 year ago (1 children)

Dude, you’ve really got a h****n for algorithms, don’t you?

[–] blue_berry@lemmy.world 1 points 1 year ago (1 children)

I don't now what you mean with h****n tbh

[–] CaptObvious@literature.cafe 2 points 1 year ago

It was rude. I apologize.

[–] ttmrichter@lemmy.world 2 points 1 year ago (2 children)

I would agree to algorithm-generated timelines if I could opt-out of them and not have them turned on behind my back like what happened constantly with Facebook (to the point I had to use a browser plug-in to force the feed I wanted!) and Twitter. If other people want to abrogate their own responsibility to curate their own world for the ease of letting someone else decide what goes before their eyes, that's no skin off my nose. It becomes a problem only when that's forced on me.

[–] CaptObvious@literature.cafe 1 points 1 year ago (1 children)

The problem is, as we’ve seen repeatedly with the corporate “social” media platforms, algorithms will eventually be turned on by default with no way to turn them off. The only rational path, to my mind, is to never turn them on in the first place.

[–] ttmrichter@lemmy.world 4 points 1 year ago (1 children)

There would be no reason for a non-commercial site to forcibly turn them on, especially in an environment with the relative ease of moving that the Fediverse offers.

Corporate sites turn them on because they're milking you for ad clicks and personal information. You make them money the more often you click things. Private sites, conversely, LOSE money if you stay there 24×7. The dynamics are different.

[–] CaptObvious@literature.cafe 1 points 1 year ago

Oh, I agree on all points. My post wasn’t clear. I meant to say that if algorithms were available at all, they would eventually be forced on everyone due to making moderations easier by reducing the number of posts that are actually seen. Hence the only rational approach being to never develop them in the first place.

[–] blue_berry@lemmy.world 0 points 1 year ago

I'm not sure because that could be a feature of some instances: we have chronological timelines come to us! Would instances without eventually win? Not necessarily, right? So I don't think that to be a necessity