Skip to main content


I've been thinking about the lack of algorithmic* feeds in the #Fediverse and wondering how much that is hurting growth and retention. Big social media spend billions on getting and keeping eyeballs, and algorithmic timelines are a huge part of that.

I think it's right to avoid it, but I worry that will just forever stunt growth. The alternative is human curators, I think, but I have no idea how to make that happen.

*Yes, I know reverse chronological is an algorithm, algo-pedants.

#mastodon

in reply to John Conway

Reverse-chronological-timeline-only is a feature for some people, but not _most_ people.
in reply to John Conway

i feel like this sorta brings us back to the question of what social media is for. is it a place to make connections with people? is it a place to sell things? is it a word-of-mouth network? an entertainment site that should keep me stimulated like a tv?

i feel like the latter one sort of doesn't fit, even tho that's the direction corporate social media has gone. i dunno. does that actually help artists sell work? hard to know if even your followers will see your stuff

in reply to no-no

@wolfteeth Well, I think it's innovation over forums is that it's all of those things. And in being all of those things it brought a huge number of people that artists can reach (and they do reach people, there's no denying it to popular artists)

Personally I actually prefer the forum format, and if I was just doing this for my own amusement/making connections, that's what I'd do.

in reply to John Conway

heheh, i do find myself thinking of setting up a forum from time to time, but i think, besides social media, discord has fully taken over the corner that used to be occupied by small forums. i sincerely like the microblog style of social media, though, so i am totally here for my own amusement 🤷

i still question the usefulness of "the algorithm" for reach. someone will probably write an app like that for fedi sooner or later, and it's not a bad idea. just not one that interests me much

in reply to no-no

@wolfteeth I mean, I'm biased here because I know how much artist rely on breakout hits to increase their viewership. Algorithms tend to amplify that.
in reply to John Conway

I feel like there needs to be a citation there. I'm totally willing to believe that limited groups of people are put off of fedi because they can't see the 'hot' content first or some other algo - but every person on Twitter I have ever seen comment on algorithmic content resents bring forced to participate. Even making that kind of thing optional poisons the pool for how people interface with the fediverse.
in reply to Bel Lion

@bel_lion What sort of citation could I provide you with?

All the huge social networks use algorithmic feeds. You may say its so they can control the message, but I say the simpler explanation is that it keeps people engaged because they like it. TikTok is a case in point. Many people love the algorithm there.

I'm not saying this is the way to go, but I don't think that just assuming people don't like recommendation algorithms is the way forward either.

in reply to John Conway

I certainly don't debate that the algorithms used by corporate social media keep people engaged with the platform. I simply am unsure that 'liking' it figures into the effect.

That's why I asked for a citation. All I have right now is my own observations, which seem to conflict with yours on this issue. I was hoping that you might have some research source I could read and perhaps understand your view better.

in reply to Bel Lion

@bel_lion Well, if the fediverse can't do anything to improve growth, the question is completely irrelevant, because people will stay on Twitter/Instagram/TikTok.

What would you do or change about the current situation?

in reply to John Conway

I think that getting friction issues like quote-toots, instance-based timeline search (not third party), and similar will do a great deal to retain new prospective users. Giving users the tools to control their own feed at the client level is great! If a client builds an algorithmic timeline feature local to the users cache, that's fine!

Where I perceive this being a problem is if such implementations happen at the instance level.

in reply to Bel Lion

@bel_lion I agree, and actually that's what I'm hoping will work. Application-level user-controlled algorithms could be great.
Unknown parent

John Conway

@Colman What's "empty" growth?

From my own selfish perspective, yes, the Fediverse is too small to support the number of creators on here. It needs to grow for that reason .

From a broader perspective, I'm afraid that hovering around a couple of million users isn't the kind of change in the internet landscape we needed. It's no where near enough to break the hold of the giants.

Unknown parent

John Conway

@artcollisions I want some societal change to come out of this, not just an okay place to hang out on the internet. The Fediverse has not yet done that.

Also, it's just not really big enough for creators to make a living.

Unknown parent

John Conway
@BrianJohnson As someone that has made a living out of using social media, there are certainly ways you could naively interpret your numbers, but the network effects are real.
Unknown parent

John Conway
@Colman I think this is a mischaracterisation of Twitter actually. The outrage part of it existed of course (and boy was I sick of it!), but there were also just a whole heap of people there to talk about/follow their interests.
Unknown parent

John Conway
@artcollisions Breaking the billionaires hold on our attention. Most of the first crop of social media billionaires were I think mostly confused by what they had made, but that will change. Social media will be used very deliberately like the old media for specific agendas (think Rupert Murdoch). We need to have broken their hold before that happens.
Unknown parent

John Conway
@Colman Yeah, but I worry they'll just make the same mistakes all over again (because they work!).
Unknown parent

John Conway

@Colman Commercial instances are a near certainty if the Fediverse becomes reasonably successful. And I worry the effects are subtle. Not sure we'd recognise it for a long time.

On the other hand, at least it wouldn't be one or two billionaires deciding, and that would be a huge win in itself.

in reply to John Conway

Honestly, I think algorithms aren't the issue. I think the issue is one of discoverability.

It's about being able to cross-server find accounts.

Meaning, of course, I think that cross-instance search is the best tool for this. Since it let's people build local feeds with exciting and regular content.

in reply to John Conway

@John Conway Your whole post hinges on growth being desirable. However, with no profit incentive on the #Fediverse, there's no need for growth either. Commercial social networks spend so much on growth and retention because it's integral to their business model, not in their users' best interest, see the toxic culture fostered on Instagram.
in reply to Hypolite Petovan

100%. Why should "growth" be important anyway? It should be about having a "social" and a "network". To follow what you like, to interact with your friends, and so forth.
in reply to Hypolite Petovan

I think the larger point about people being able to build local networks on the Fediverse, which they can use to create.

I understand that creating on the internet is a major business thing these days, and I don't know if I think it's worth burning down a network for, but I think that it's something to consider.

in reply to silverwizard

@silverwizard Even through an artist point of view, regular social media algorithms don't do much for them, mostly hiding their content from some of their followers, and increasing exposure inequality by recommending already popular posts/artists.
in reply to Hypolite Petovan

Sorry, yes, I wasn't saying the algo was good. I believe John's point is that algos drive critical mass of users, which is important for artists &c.

I think John is not fully correct, but I think the lack of stickiness slow things down

in reply to Hypolite Petovan

@hypolite I think it might be useful to have more nuance in the terms we use here. One type of growth that I think is important is people being able to keep their *social graph* and that requires a certain *critical mass*

I moved to the fediverse full time and tried for years to bring people here from facebook and the #1 complaint I got was "you are the only one I know that uses it"

a large portion of people use social media to interact with already existing friends not make new ones.

in reply to wakest ⁂

@hypolite We don't care about growth for the sake of growth. But social media is only useful if the people you want to interact with are using it. Otherwise its just a place to take notes into the void and thats not much fun for most people...
in reply to Hypolite Petovan

@hypolite In my opinion, there's little point to the whole thing if it doesn't grow to address the problems with current social media.

We may as well be hanging out on a big forum (which is fine if you want that, but that's not why I'm here).

in reply to John Conway

I've been thinking about this too, and how I personally would like some kind of optional 'in case you missed it' algorithm that prioritized posts from people I have a history of liking or responding to, since I've found I miss those often in the pure-chronological timeline
in reply to Henry Meyers

@mojoceratops Yeah, I agree. But I think there's a danger in "or responding to", because that can lead you to seeing people you argue with too often (this happened to me on Facebook, and made my feed an irritating experience).
in reply to John Conway

yeah good point, might need to limit to likes only to avoid that feedback loop
in reply to John Conway

What you call stunted growth, I call organic growth. People come because someone recommends it, stay because they see things they like, don't leave because it's not toxic. This is all good. Slower growth gives time for server admins to scale up and find funding strategies. Exponential growth gets you where you want to be soon enough, even for a small exponent.
in reply to John Conway

That's fine: it's tourists who joined in the initial burst, and never really engaged. They're free to go.
in reply to Mike Taylor 🦕

@mike Yea, I'm hoping its a slump after and period of false growth, and real organic growth will return. The truth is though that the Fediverse is just too small for the palaeoart community for example. Growth is needed.
in reply to John Conway

To small for the palaeoart community to *what*? I'm seeing more actual communication between you guys here than I ever did on Twitter.
in reply to Mike Taylor 🦕

@mike Support an audience that leads to Patreon conversions, book sales, commissions, etc. It's mercenary, but there you are, it's how we make a living.
in reply to John Conway

OK, you're talking here about a broadcast medium?

(Nothing wrong with wanting that, of course; but I think it's a mistake to want THIS to be THAT.)

in reply to Mike Taylor 🦕

@mike I don't agree that treating it something like a broadcast medium is a mistake. Twitter and Mastodon are *explicitly* asymmetric to allow this, unlike Facebook where the relationships are symmetric.
in reply to John Conway

Interesting. And yet Mastodon is explicitly non-algorithmic. And Facebook's weirdly named concept of "pages" are explicitly asymmetric for just this reason.
in reply to Mike Taylor 🦕

@mike Yes, Facebook added a specific mechanism for it (because Facebook was trying to be an everything app).

I'm not sure where in this conversation everyone got the idea that I was arguing that algorithmic timelines are a great idea – I explicitly say the opposite in the first post – my point is that Mastodon competes with big social networks, and algorithmic timelines are a competitive advantage for many, many people.

We need to give them something competitive.

in reply to John Conway

What, other than algorithmic timelines, would be competitive?

Also: why would algorithmic timelines delivery you more potential-customer eyeballs than chronological timelines?

in reply to Mike Taylor 🦕

@mike Algorithmic timelines are actually two things mashed together:

1. A filter, prioritising the most popular posts from the people you follow, which is meant to keep the entertainment value high

2. A recommendation engine, which is meant to aid in discovery

We can probably do without the first. The second gets to the huge discoverability problem on Mastodon. People don't know who to follow or how to find stuff that's interesting to them, and they often give up.

in reply to John Conway

I am certainly 100% in agreement that discovery is very poor.

I think the answer to that is search, but we've already discussed the short-sighted cultural factors that have so far prevented this.

in reply to Mike Taylor 🦕

@mike I'd like to see search and a recommendation engine API of some sort. The engine would have to be _very_ carefully built though.
in reply to John Conway

I'm not sure what recommendations would get me that searches didn't. But I have no argument against it being tried so long as it's strictly opt-in.
in reply to Mike Taylor 🦕

@mike You have strong interests and intent. A lot of people just don't, they use social media for light entertainment, and that's fine by me.
in reply to Mike Taylor 🦕

@mike I'm thinking more of a "StumbleUpon for the fediverse", rather than a "silently alter your feed based on feedback you didn't even realise you were giving".
in reply to John Conway

There are opt-in Fediverse full-text search sites for those who want to participate. The number of participants on those who fully open their posts to be searched by others, though (as opposed to just themselves) seems to be very very low. The one site I know allows relatively granular control over what others are allowed to search on and seems to (from my initial look) to be mostly locked down by most users to just themselves. I just offer this as a data point in the discussion of search for the Fediverse.
in reply to DannekRose

@dannekrose @mike See, I've been here a while, and I don't know what they are, or how to opt in. This probably needs to built into Mastodon mainline to work.
in reply to John Conway

@dannekrose Strong agreement on this. Nudge Theory and all that. People mostly only see what's in front of them. Me included.