Algorithms aren't the enemy. Chronological feeds don't scale and the signal-to-noise ratio will plummet if this ever gets popular. The real problems with today's algorithmic feeds are non-transparency, lack of choice, and optimizing for engagement instead of healthy discourse.
Open-source is a perfect opportunity to fix all this. Have there been any efforts to create a Mastodon instance with a (community governed) ranking algorithm? Is that technically feasible? Or is the idea simply anathema?
Responses to some frequent comments:
* I'm certainly not suggesting that algorithmic feeds should be imposed on everyone! Choice is great. I recognize that many, perhaps most current Mastodon users like chronological feeds.
* "Reverse chronological" is an algorithm, albeit a simple one. It's currently the only option. Chronological feeds are not normatively neutral. There is, unfortunately, no neutral way to design social media. https://mastodon.social/@randomwalker/109308664849924122
* "Mastodon doesn't need to become popular." Sure. But like it or not, it's getting more popular, and many of the newcomers have a different culture and expectations. Eugen Rochko: "People who are arriving now have as much right to be here and bring their own culture as the ones who came before them." https://mastodon.social/@Gargron/109323118267580967
Again, all I'm suggesting is choice, and I thought Mastodon is all about choice.
* What do I mean by chronological feeds don't scale? A few things.
1. There's a lot of social pressure to follow people (especially people you know). Old-timers here are comfortable with following a small set of people, but most newcomers aren't. For those people, pretty soon the feed becomes a firehose.
2. Even in a mostly-chronological feed, some ranking would be really nice. *Not* necessarily by popularity, but if there are 15 posts by the same person I don't want those to be the first 15.
Amusingly, this thread is getting boosted a lot today, a couple of weeks after I first posted it. That's another difference between chronological feeds and Twitter's algorithm (which heavily emphasizes immediacy).
But—and this is the point of the thread—imagine all the interesting things you could do with a tunable algorithm. You could even let the recency preference depend on the type of content! You could customize it to show you news only if recent, but educational content regardless of age.
@randomwalker These are important points, and I agree with most of them. My Twitter engagement was better under reverse-chronological sorting, but you rightly point out that that's still suboptimal.
When I say "the algorithm" for Twitter, I mean the machine learning recommendation engine. Reverse-chronological is an algorithm, but it's deterministic.
Having worked in various AI-related fields for years, I think it's important for the algorithm we use here to be predictable and understandable.
@randomwalker hi! Reverse timeline fan here -- not bc it's the best, but bc the alternative was much much worse in my case.
You're suggesting users should tune the algorithm. Fine.
On twitter, the algo tunes the user. Unacceptable.
The solution is getting back to personal web sites
Aggregation and curation would be done by hand, by people in telegram chats
@randomwalker not sure if you’ve seen it, but the renewed interest could also be because of this useful thread and discussion from @jon, an ex-Twitter designer
4/ I remember when we moved from strict reverse-chron (which 97% struggle with) to "You might like" prompts (which 3% struggle with) and hearing from the VERY LOUD minority that we were destroying Twitter. But we saw as the 97% had a much better time. We saw that every step forward we took (I have a whole presentation on this) was helping people more and more. The data told us we were making a better product. And that reverse-chron kinda sucks. For most people.
@randomwalker We can certainly imagine a user-tunable choice algorithm, and try to make its decisions interpretable.
Have no idea if current architecture allows a feed algorithm, or if it would be extra work.
But I don't think any idea is anathema in federated land, things are just more or less work for devs doing this mostly-unpaid labor. (Thanks devs!)
Do you mind elaborating on why you think "sorting and organizing misinfo if a model goal is a healthy community." ?
@goabiaryan @riedl @randomwalker
I can sorta see that, but I also did like having a way to add context to a recommendation. A lot of science-y QTs added background information, like explaining why a thread might be interesting to a broader audience.
It could be user-configurable, i.e.
◻️ Allow my posts to be quote-boosted
◻️ Don't allow my posts to be quote-boosted
◻️ Let me decide for each post
While quote-tweeting is sometimes toxic it also often drives discussion/engagement and it doesn't seem optimal to not have the functionality in some form.
@mil @randomwalker Does noise to signal ratio improvement scale without personalisation? I think scale suggests bigger is better, maybe not?
Mastodon allows some #control around its features at the #community level like maximum message length I hear. I often wondered if the length of the messages on twitter was a problem. Lots of the best stuff I read was in long and annoyingly formatted threads.
It's a free world :). There are interesting forks and implementations with different behaviors, such as Hometown, which emphasizes local posting:
https://github.com/hometown-fork/hometown
Personally, I'd love the ability to temporarily turn the volume down for some frequent posters, or to have their stuff collapsed when the post velocity gets very high - that'd be enough for me at least at current scale.
@randomwalker Good question. On Twitter, I tended to follow a LOT, gave up on seeing everything in the timeline, relied on the algorithmic timeline, and aggressively used filters.
I'm not sure how to adapt that approach here, or even if I should.
@randomwalker I'm imaging an alt-timeline view that amounts to "show me toots from my home timeline (or local server or other?) over the last $TIME_INTERVAL, ranked by $CRITERIA, which may be evaluated over local context (boosts, keyword/hashtag-assigned bonuses or penalties, etc.)
Unsure if I would need deeper context (e.g. level of prior engagement with authors of those toots).
@randomwalker Also, more generally: please spend some time in the fediverse and get to know the local culture first, before suggesting sweeping changes that directly go against what people are here for...
If people here wanted things to work like on Twitter, we'd be on Twitter
@randomwalker But wouldn't the loudest & most frequent speakers then control the WHOLE conversation? If someone is on here all the time, always tooting and always ranking, they would have significantly more impact on what everyone's feed looks like than the person that comes on once a day or so.
Just create some lists with your favorite posters. Mute over-posters. There are plenty of ways to customize your individual feed's signal-to-noise ratio without pushing algorithms on everyone.
@randomwalker Algorithmic timelines are only evil when it’s the only choice or the default choice. People should be able to choose which one they prefer.
Both Twitter & Tumblr give me the option to choose between an algorithmic timeline or a chronological one.
Tumblr goes several steps further by suggesting sites based their recommendations, hashtags I follow, blogs I follow, trending & popular reblogs.
But both offer a simple chronological feed which is awesome.
@darnell @randomwalker I think you both might be missing the fact that the other folks on your server act as moderators for your local and federated feeds.
So if you join the right instance, 'local' becomes a human curated equivalent of the algorithmic feed.
That starts to break if everyone just uses the AI.
@darnell @steely_glint @randomwalker There thing about OSS is having the choice.
I like to follow lots of people with diverse views. A reverse chronological algorithm means people who post infrequently will likely be lost in the noise. I wouldn't mind an algorithm that added some weight to their posts so I see them.
@randomwalker I think there are tight limits to how much #algorithm #choice will help because the vast majority of people rarely configure or change any settings on the apps they use, never mind algorithms. In fact a choice approach could become a way to individualize the problem and skirt #accountability.
Perhaps the algorithms aren’t the enemy, but in that case, we need to address the people who choose to engineer and deploy default algorithms implementing harmful but profitable #incentives.
David Kobia
Arvind Narayanan
Aranjedeath
Swen
LiamSora 
Dr. Angus Andrea Grieve-Smith
politiconomics
uhmmm
Toby Phillips
rival
Darnell Clayton 
Jesse
D. G. Fitch
Seven
Michael Ekstrand
Giovanni Beltrame
Nathan TeBlunthuis
Taylor Beauvais
Brian C. Keegan, Ph.D.
Mark Riedl
Abi 
Matt Krause
Barry Devereux
Leena
le Pétomane Ancien
🚲
josh buermann
Dan
Erik Moeller
Uncomics
Sean Blakey
Denny Vrandečić
Sven Slootweg
orko
SLaparle
Tim Panton
Atz
Andrew Hundt
Philipp Leitner