Mastodawn

#Bluesky now allows you to “choose your own algorithm”.

Which sounds “incredible” and “sci-fi”—but it really isn’t.

What it essentially does is give a Twitter-like service Reddit-like features.

As an aside, now I’m wondering why Reddit doesn’t offer an alternative web front-end to make it more Twitter-like!

But how does this apply to the #ActivityPub flavour of the #Fediverse? This feature now makes me realize how big a deal Fediverse groups are going to be, and if I were @Gargron, I’d be even more excited about rolling out Mastodon’s group functionality.

Because while groups aren’t exactly relevancy algorithms, once you add a “New”, “Hot”, “Best”, etc. feed to groups, now you’re in business.

I don’t know if choosing your own algorithm is the killer feature that Bluesky thinks it is. My experience is that most people hate choice. Nevertheless, I still thinks it’s important.

@[email protected]

Chris Trottier May 29, 2023

If this is "choosing your own algorithm", #Calckey has actually had this feature long, long ago.

We call it "Antennas" -- and you can easily build an Antenna yourself.

Here's a screenshot for Antenna creation settings.

Chris Trottier May 29, 2023

So I'm thinking about this in more detail, and I think the "choose your own algorithm" feature with #Bluesky is really not hard to implement, and something very doable on the #Fediverse.

What do we exactly want from an algorithm? Topics.

And we want topics sorted according to the following:

1. Hot
2. New
3. Top
4. Rising

Some people would like a "controversial" feed, but we don't have to give it to them 😉

As for "Top", we can sort it according to time parameters.

Mike Fraser May 29, 2023

@atomicpoet I've always envisioned it as something more personal. Like something that would go through your incoming feeds, pluck stuff out it thinks you'd like and present it in a curated feed. I was in a discussion about algorithm bots. Basically a bot that would go searching for your criteria and repost it. You could then throw it in a list or whatever. Algobots roll out!

SparkIT May 29, 2023

@atomicpoet I think #kbin magazines are also a good example of something similar to Bluesky's lists because they seem to work like subreddits but every user can create one choosing several hashtags to follow (and maybe other conditions? Not sure about that) and other people can subscribe so it's an interesting mix.

Chris Trottier May 29, 2023

@sparkit I was actually thinking about this too.

ᗰᗩᖇǤᗴ against Putin #NAFO May 30, 2023

@atomicpoet I don't know if this belongs here. I find it unfortunate that current events like earthquakes, hurricanes, etc. take a long time to arrive in the Fediverse. Can something like this be improved by an algorithm?

♾️ Yuki (스노 雪亮) 🐬 🧮🗝️May 30, 2023

@aniho91

I think it relies on various factors:
1. Is the account pushing emergencies and disaster alerts a bot sourcing from Twitter or elsewhere? Or, is it an official account?
2. If it is an official account, are they posting it manually or automated? If it is the latter, how do they prioritise posting to the fediverse?
3. The fediverse instance they are using, did that instance set a reasonable setting for pushing to other servers?
4. If so, is the instance's software not throttling down the account in any way?
5. For your instance, how frequent are they polling the accounts you are following? Is it reasonable? Or, is there a time period when the load is high that it slows things down to make it manageable?

There are probably other factors involved.

I think the best way is for the fediverse software to have some flag for verified emergency/disaster alert accounts which will prioritise them in pushing their posts. However, the receiving instance would have to know which accounts are official for them to be able to increase their priority on their end as well.

@[email protected]

ᗰᗩᖇǤᗴ against Putin #NAFO May 30, 2023

Now I know I know nothing about algorithms.😳

cc: @atomicpoet

Steve Bate May 30, 2023

@atomicpoet How would you determine "Hot", "Top", "Rising"? Mastodon, and probably other implementations, don't federate likes. That may be true for reply counts too.

Chris Trottier May 30, 2023

@steve For trending posts, Mastodon uses a combination of likes and boosts.

Steve Bate May 30, 2023

@atomicpoet Yes, but that Mastodon "like" data is instance-local, right? I'm not sure if boosts are only published to followers or not. In general, I like the idea but it wouldn't work well for single-user or small instances like mine.

Jupiter Rowland May 30, 2023

@Steve Bate It doesn't?

If I like a Mastodon toot on Hubzilla, my like can be seen on Mastodon. Vice versa, if someone on Mastodon likes one of my Hubzilla posts or comments, I can see it. And I'm pretty sure that all this doesn't only happen on the side of Hubzilla where ActivityPub is an optional add-on.

Steve Bate May 30, 2023

@jupiter_rowland For example, one of @atomicpoet recent posts currently has 19 likes and 9 boosts. It displays as 0/0 on my instance. I've read that Mastodon doesn't federate "likes" because it would be too easy to spoof and some users don't like exposing that data outside their local instance. I'll look for a reference.

Nasos Alaiskas May 30, 2023

@steve @[email protected] @[email protected] that is a true question, that I would like to be answered/solved.

eshep May 30, 2023

@atomicpoet @steve, could you expand on "don't federate likes" please?

Gustav Lindqvist May 30, 2023

@eshep @steve @atomicpoet IIRC they are sent one way from the one doing the like to the one receiving it. The one receiving it does not broadcast it out to everyone else. So you can see a pretty good count for likes on your own posts and you can see your own likes on others posts but everything else is miscounted.

Steve Bate May 30, 2023

@gustav @eshep @atomicpoet That makes sense. Thanks for the clarification. My point is that this signal can't be reliably used for timeline ranking, at least in smaller instances. I'd personally like the ability to use custom ranking based on post data like content, author, instance, and so on (could include local likes too). I'm thinking of something like a Bayesian or ANN boost/filter trained with my own historical likes and/or boosts.

Allen but one of the good ones May 30, 2023

@eshep @steve @atomicpoet From what I've seen, if you and I are not on the same instance, and we both see a post from somebody on a third instance, and we both like that post, then the following happens:
OP sees both our likes. But I don't see your like, and you don't see mine, unless we use "View on original page"

eshep May 30, 2023

@steve @atomicpoet @uastronomer, hmm, hadn't noticed that. Do you see any likes on this thread from my question to here? I see only the two I just made.

Allen but one of the good ones May 30, 2023

@eshep @steve @atomicpoet Nope! The only like I see is one I made on Steve's "How would you determind..." post a bit further up the thread.

eshep May 30, 2023

@steve @uastronomer @atomicpoet Thanks, good information to know! I don't see any likes on that post. Now I'm curious how that ""feature"" works across other softwares. I wonder if it's just another case of Mastodon not playing well with others, or this is equally shit across the board. Anyone tested this thoroughly between different AP interfaces?

Allen but one of the good ones May 30, 2023

@eshep @steve @atomicpoet I could be wrong, but I seem to recall this being a design feature of Mastodon itself - the idea being that "Going Viral" is everything bad about social media, because disinformation and chasing clout and so on, so the idea is to hide a post's popularity so that people are less likely to respond to a message because of it's popularity and more likely to engage based on the content of the post itself.

Personally I think that this is just a feature of how human brains evaluate the importance of information, it's a feature of how we communicate, and that it is how ideas naturally propagate through societies and cultures. I think that social media merely amplifies this process, and did not create it from nothing, and that trying to solve all those problems by amputating it completely is a mistake that ends up neutering conversation completely... but what do I know, I'm just some random guy on the Internet *shrug*

eshep May 30, 2023

@steve @uastronomer @atomicpoet Good point with the "viral" garbage. I don't dislike that likes wouldn't be seen by anyone other than the two parties directly involved. It makes sense when you translate it to real life conversations. Usually, people don't advertise how many people agreed with something they said, or how many nodded their head when a good point was made during conversation. But as you said, forcing people to not do things they would prefer to otherwise, never goes as expected.

DuncanMSussexPol May 31, 2023

@atomicpoet I agree this is something #Mastodon and the #Fediverse in general is lacking.

Back on old #Twitter I really didn't like the introduction of algorithms when that happened but having them on sometimes did improve the experience.

I guess one thing that would be good on top of those would be perhaps a "slim" chronological feed. The biggest problem with chronological is when you follow both very active and very inactive accounts, a bit of weighting would be good.

Rumblestiltskin May 29, 2023

@atomicpoet That is definitely an algorithm but I'd like to see an algorithm where likes and retoots can influence the order the they appear on the timeline. Maybe even the likes and retoots can be weighted based upon the datetime the likes and retoots were done.

I think that would that could create some interesting results.

Gabriel May 29, 2023

@atomicpoet

Is there a way to filter by language?
I can't find anything in the settings and I can't think how to do it in the antennas.

Kainoa May 29, 2023

@gabriel @atomicpoet There's no language filtering anywhere in Calckey yet -- rest assured, it's a planned feature.

Lyre Calliope May 29, 2023

@atomicpoet can antennas be shared?

Miss Koula May 29, 2023

@atomicpoet great but buggy feature... Now working the most of the time...

@atomicpoet this might be a reason for me to move to calckey

joel b May 29, 2023

@atomicpoet @Gargron @fediversenews but what if I choose the wrong algorithm? It would be more difficult then moving to a new neighborhood to change! /s

Onion May 29, 2023

@atomicpoet @Gargron @fediversenews
I totally agree. I don't like it if there is too much choice. But at the same time I think it would be great if there would be the minimum choice to activate or deactivate some kind of standard algorithm and then, if you want to, you can modify it as much as you want if you have the time to get into how to do that.
Maybe there is something like that already here that I don't know about.

Matthieu May 29, 2023

@atomicpoet @Gargron @[email protected]

I don’t like algorithms much, because every algorithm becomes a game and some clout-chasers are very good at games and even cheat. The Bluesky’s "12 likes to be on what’s hot" feed was already pretty terrible. You just need 12 friends to get in there. The new "what’s hot" is also terrible, because it’s a reddit-like hot algorithm except for the entire platform not for a group and once a post get there it can stay on top for the entire day. It’s basically reddit karma farming. 0% chronological (But understand why people with large following like this)

I also don’t like the idea of likes becoming a measurement of a post’s worth. On mastodon likes are benevolent and boost is the actual "I want more people to see this" button. What about emoji reactions? If I react with a 🫠 is this supposed to be a like? A dislike? To me the worth of a fediverse post is by the number of boosts and I hope it stays this way.

What I PARTICULARLY don’t like is the way bluesky implemented their custom feeds. It’s not a plugin written in AIscript running on your instance server nor on your client. It’s literally some random dude’s server running unknown code. Meaning on top of trusting the software running on your client, the software running on your instance server (PDS), the software running on the BGS (which is a separate service) you know have to trust N services offering you N different "custom" feeds. "Make your own feed" doesn’t mean select a few keywords, it means run an entire server on your own.

But still, we can learn from bluesky. Algorithms aren’t all bad and we can have more than just chronological orders. The Mutuals feed is very popular and simple to add, the catch-up feed is interesting when people login only once a day and don’t want to scroll their entire timeline since yesterday, etc. We can add a little bit more choice in our timelines.

But I still I prefer my antennas than having to ask a third-party dev, please can you make a custom feed following this keyword for me? This isn’t freedom of choice to the users...

Chris Trottier May 29, 2023

@matthieu_xyz @Gargron @[email protected] This is a good summary, and explains everything well. My takeaway is that we can do it better with less resources and with more transparency. Antennas really are a better option.

narF 🎲May 29, 2023

@atomicpoet @fediversenews @matthieu_xyz I wrote this thread about sorting algorithmn for Mastodon 3 years ago and it got a lot of interest and replies.

https://mstdn.fr/@narF/104887310401243826

It's sad that there isn't an implementation of it yet.

narF 🎲 (@[email protected])

I want to talk about algorithms on social media. Specifically about how mastodon and the Fediverse often write on their presentation page or feature list about how they "don't have algorithms" because they display posts in chronological order. 👇A thread

mstdn

narF 🎲May 29, 2023

@atomicpoet @fediversenews @matthieu_xyz Can Antennas on Calckey do what I'm describing in that thread?

Matthieu May 29, 2023

@narF @[email protected] @[email protected] This is what the antenna creation UI looks like. It’s nothing like a bluesky custom feed that can detect pictures of cat with AI. But it’s much stronger than just following a hashtag.

So strong that users are even afraid of people finding their post when they thought they were relatively hidden and hard to find.

btrandolph May 29, 2023

@matthieu_xyz @atomicpoet @Gargron @fediversenews

the heat is too hot

Laurens Hof May 30, 2023

@matthieu_xyz @atomicpoet @Gargron @[email protected] good post, I agree.

I do like the feed that just got added today on bluesky, that shows the first post of new users signing up. That might be helpful for onboarding new people, and something that fedi can probably take some inspiration from.

Chris J. Karr May 29, 2023

@atomicpoet @Gargron @fediversenews

When I get some time, I'll be looking into implementing an algorithm that I can apply to boosts where I can apply a positive, negative, or neutral rating to boosts from particular posters and have those surface (or sink) in my timeline based on the ranking score of the booster and boostee. I'm also looking into a "seen before" filter to skip the duplicate boosts as one goes viral.

Running a server for one, so I'm not looking to apply this to others' feeds.

ᗰᗩᖇǤᗴ against Putin #NAFO May 30, 2023

@gavinkarlmeier @dennishorn FYI “eigenen Algorithmus auswählen“

maegul May 30, 2023

@atomicpoet @Gargron @[email protected]

What it essentially does is give a Twitter-like service Reddit-like features. I think I’ve been saying the fediverse needs this since I got here.

Once you take away a shadowy manipulative engagement driving algorithm, and rely entirely on reverse chron and your social network, there’s a lack of structure that is socially off putting and exclusive.

Fusing Reddit like structures with an algorithm free social network can add discoverability and engagement to a social connections first platform/ecosystem.

Topic based structures help you find people, and filtering/sorting features (which I think is a better name than algorithm which clearly now connotes manipulation) help you find what everyone is talking about so you don’t feel left out, while leaving you in control.

I hope federated groups become normal and take off as a major way of interacting in the fediverse, even to the point that instances become less significant.