My Problems with Mastodon
Even with growing pains accommodating an influx of new users, Lemmy has made it clear that a federated social media site can be nearly as good as the original thing. I joined Lemmy, and it exceeded my expectations for a Reddit alternative run by an independent team.
These expectations were originally pretty low when Mastodon, the popular federated Twitter alternative, was the only federated social media I had experience with. After using Lemmy, Mastodon seems to be missing basic features. I initially believed these were just shortcomings of federated social media.
-
Likes aren't counted by users outside your instance, and replies don't seem to be counted at all (beyond 0, 1, 1+), leading to posts that look like they have way more boosts (retweets) than likes or replies:
This incentivizes people to just gravitate toward the biggest instance more than people already do. My guess is that self-hosting a mastodon instance would also not be ideal, since the only likes you'll see are your own.
-
There's really only one effective ways to find popular or 'trending' posts. There's the explore tab which has 'posts', and 'tags' sections.
The 'posts' section shows some trending posts across your instance and all the instances that it's federated with, this is the one I use it the most.
The 'tags' section is a lot like the trending tab on Twitter, but it's reserved just for hashtags, which I guess isn't a huge deal, but it feels like a downgrade. However, I do like the trend line it shows next to each tag!
The 'Local' and 'Federated' tabs are a live feed of post from your home instance and all the other instances, respectively. I feel these are pretty useless and definitely don't warrant their own tabs. Having a local trending tab for seeing popular posts on your instance would be more interesting.
-
The search bar basically doesn't work, is this just me???
-
This one is more minor and more specific to a Twitter alternative, but when looking at a user's follows, you'll only see the one's on your home instance but for some reason this rule doesn't apply to followers.
From what I've heard, a lot of these issues are intentional in order to create a healthier social media experience. Things like less focus on likes, reduces a hivemind mentality, addiction, things like that (I couldn't find a source for this, if anyone has one confirming or disproving this please lmk).
Why this is a Problem
Mastodon seems to have two goals: To be an example of how a federated alternative to Twitter can work well, and to be a healthier social media experience. It's not obvious, but I think these goals conflict with each other. A lot of the features that are removed in the pursuit of a healthier social media will be perceived as the shortcomings of federation as a concept.
In my eyes, Mastodon's one main goal should be proving federated social media as a whole to the public, by being a seamless, familiar, full-featured alternative to Twitter. For me, Lemmy has done that for Reddit, upvotes are counted normally, you can see trending posts locally and globally same with communities, and the search function works! All its shortcomings aren't design flaws, and I fully expect them to be fixed down the road as it matures.
As annoying as Jack Dorsey is, I have high hopes for BlueSky.
Likes aren't federated well on Lemmy either. Posts can have wildly different scores on different servers.
No content algorithm is kind of the point of Mastodon. That was a conscious decision. There are trending hashtags you can follow, though. There's nothing preventing someone from making a Mastodon/Misskey/Akkoma fork that will come with its own popularity system, but you'll have to find (a team of) developers that care about such things to get it off the ground.
The search bar works fine for me, maybe your instance is overloaded?
The lack of back fetching is annoying, but it's impossible to do without causing massive server load. The push-only ActivityPub implementation is a lot lighter on server resources (which are already very high for Mastodon for reasons I'm not entirely sure about). Fixing this would break or slow down Mastodon servers across the network. It should be noted that Lemmy suffers from the same problem, though it tries a little bit harder when it subscribes to a remote community for the first time.
I don't think Mastodon wants to be "another Twitter". I'm pretty sure they want to be better than Twitter, with their own definition of better. In my opinion, and seemingly a lot of the Fediverse's, that means getting the stuff you're subscribed to and trying to hold some kind of popularity contest.
It you do go by popularity, there is absolutely no way of detecting voting rings and boost bots. There was a recent Lemmy example of someone using their own instance to send thousands of upvotes, bringing their post about voting rings to the top of every sorting method. Popularity works with systems that are centralised because bot detection can happen in a centralised place, gatekeeping the content behind an anti abuse wall. With federated servers, you will need to choose between "whitelisting every server you federate with after vetting their bot detection" or "accepting that your popularity feeds will soon be full of spam". Neither is a great option, in my opinion.
My take on your analysis: you seem to be on a huge Lemmy server and on a small Mastodon server so you don't notice that the same class of problems exists here too.
Bluesky solves the federation problem by not being very federated. Their idea of federation is "everybody downloads a full copy of the network" which is unsustainable for any independent server. They see the future as "a bunch of Twitters talking to each other, each with their own content recommendation and filtering mechanism" rather than "something anyone can become a part of for a minimal fee".
Im glad you pointed out the algorithm thing. Seems like people get fed up with social media platforms like X(?) and Reddit and then come to alternatives demanding the same features that, at least in part, led to them being fed up in the first place.
I actually disagree with OPs assertion that these federated platforms are 'almost as good'. Theyre better. More features doesnt mean a better platform and in my opinion often makes them worse.
I think it makes sense. Algorithms like the ones that Reddit and 𝕏 use are what keeps users coming back. They lure you in with addictive content so you keep coming back. People have started using social media for entertainment.
This is probably why so many people are turned off by Mastodon for very vague reasons. Often they'll say "it feels death" despite having a healthy feed of random posts and hashtags on their instance's homepage. It's not that there's no content on the Fediverse, people have just started expecting social platforms to provide them with their daily dopamine hits, and alternatives that don't do that feel weird and disappointing. That's what I felt for a while when I started using Mastodon, anyway.
I don't want a fancy algorithm, I just want to see the popular posts from the communities I follow
Now, that's not that simple either, since popular from a big community is different from popular from a small community, but still
Honestly, Lemmy and Kbin are better set up for that purpose. Determining popularity in a way that's not easy to game requires a fancy algorithm. Bots and voting rings are easy to set up. Lemmy's sorting has been criticised for pushing ancient posts to the front page and preventing that is very hard (especially without making the server software expensive to run).
Everyone has a different definition of "popular". You probably don't want stuff that gets loads of replies because it's offensive or a terrible take to pop up everywhere, so to make sure you only get good posts, you need to do things like sentiment analysis of replies you base your popularity algorithm on. You probably also don't want to hide outcries/controversy about important people and topics, so you need to match that with the sentiment of the topic at large, and from there the list goes on.
As a workaround, you can browse hashtags. Hashtags will usually get you accounts you can follow (and add to one of your topic specific lists). Get a few active accounts on a topic and the network effect will get you tons of popular posts through boosts and replies, and suddenly you've created your own algorithm!
Or you could do the most upvoted post minus its age, like a lot of sites still do. AFAIK hacker news does nothing fancy to its "algorithm"
That's what Lemmy does! Well, sort of. The orange reddit algorithm is very similar to the Lemmy algorithms. In my experience, Lemmy is slightly better than HN (fewer posts that end up on top without any comments), but both only work when there is much more content to rank than there are pages to fill.
I think Reddit's algorithms are much better for tricking people into staying around (for me, anyway). That approach doesn't work for something like Twitter/Mastodon that doesn't really do "voting". Almost all toots have little to no interaction, even the good ones. Twitter's algorithm has been documented and large parts (but not the important weights) have been open sourced, but I don't think that can be implemented well on Mastodon.
I have found Mastodon still does that. And it turned out to be a problem, actually. I just kept going on there for no reason and reading like 100 nothings.
The original federated service is IRC and is still perfect. :)
IRC is great, if a little underground these days. It's also trivial to run your own although federating requires cooperation from both ends so it's not quite as networked as lemmy or mastodon.
adjusts glasses one could argue email is essentially a precursor to what we call federated services now, and it works as well as it always has. Predates IRC :)
This is a great analysis, thanks for compiling such a comprehensive response.