GlyphThere's been a lot of fretting recently—some of it from me—about there being Too Much Politics On Here, encouraging people to (A) stop posting so much about it and (B) post about other things. This is a valid concern but I want to take a moment to put a structural lens on this problem.
We don't have an (sigh) "algorithm" here, so we think of ourselves as liberated from the corrosive engagement-maximizing forces of social media. There is *some* truth to that, but in typical "humanity was the real monster all along" fashion, it undersells what these ML models are actually *doing*.
Twitter, Instagram, et. al. are maximizing engagement by showing you *more* of the kind of posts that'll keep you on the site longer. The kind of posts that distract people, deactivate our collective prefrontal cortex, make us anxious, and generally want to push the lever to get another pellet.
Opting out of the turbo-charging aspect where reactions beget more reactions does *not* opt us out of the fundamental truth that the anxiety-inducing, identity-affirming, mass-audience posts are just going to draw more attention and more approval. We want approval. This is human nature.
Case in point: if I write a long-form blog post that takes me 8 hours to write, that reflects my core competency (Python programming) I will usually get 100 or so 'favs' on it. An anxious (but pithy) post about a recent political development that I am peripherally aware of might do 10x better.
It takes me less time to write, it takes you less time to process, more of you will find it relevant, the whole reward system optimizes for this result. And, even here, there are structural things that reward this as well, although they apply in different degrees to different people.
If I want to take some pictures and post about finding joy in natural beauty, I'm going to get yelled at failing to put in alt text. The website will even yell at me automatically now, to prevent me from posting in the first place. Guess I'll just vent about anxiety instead.
If I want to post links to long-form writing elsewhere, I get yelled at about the "hug of death", or (if I were using a slower webserver or a static website, shout out to Twisted and Pelican) get hugged to death myself and spend a bunch of time on sysadmin bullshit.
The easiest response to getting yelled at for posting stuff people *don't* like is just … not posting. Drift away from the site, avoid things that are uninteresting to your followers, or even anyone who might see your stuff out of the corner of their eye on the "Explore" tab. Go use bsky instead.
(Shout out to @mattblaze here who has enough reach to get yelled at for just posting things that are not precisely matched to some random person's aesthetic preferences. The people getting mad about black & white photos *existing* are a real socioemotional puzzle in their own right.)
Applying stop-energy to objectionable things is not enough, if you want this to be a fun place with interesting and thought-provoking posts. You need to seek out stuff, not just the stuff that you immediately respond to, but the stuff that *you want to see more of*, and fav it and boost it.
We talk a lot about the negative aspects of (sigh) "algorithms", but *this is a thing that "algorithms" can do too*, and in the absence of such a tool, we need to do it ourselves; it will not happen automatically, because there's no automation to do it.
Be the boost you want to see in the world.
Next time you find yourself looking at a post from me that savagely dunks on some detestable scumbag and get a dopamine hit and think "get 'em glyph!!" instead of reaching for the boost button, scroll right over that and find a link to my blog or a PyPI release or something and boost that instead.
⚯ Michel de Cryptadamus ⚯@glyph as far as "being the boost", i built a customizable algorithm for your mastodon feed... kind of the best of both worlds IMHO. hard for me to look at mastodon without it at this point.
⚯ Michel de Cryptadamus ⚯ (@[email protected])
Attached: 1 image @[email protected] i built on #fedialgo to make a customizable feed algorithm that is pretty much how i interact with #mastodon these days. it's available as an #npm package. here's the demo app: https://github.com/michelcrypt4d4mus/fedialgo_demo_app_foryoufeed this is what the demo app looks like: #CustomFeeds #FediDevs #FediverseCustomFeeds #FediverseFeeds #FediverseUX
@cryptadamist the UX nerd in me flinches just a little bit at that collection of sliders and its presentation but the idea here — giving users direct control and agency over their *own* algorithm, and using an *actual explicable algorithm* and not a huge inscrutable neural network full of training data of dubious quality — is fundamentally incredibly important to the next generation of social media and I really hope it gets some traction
@glyph what you're looking at is just the demo app for the library. while i personally actually do use it to peruse mastodon it's not really meant to be a user facing situation.
the internals are implemented as a node library. there's been some discussion of integrating it into the standard mastodon web experience.
* fedialgo library: https://github.com/michelcrypt4d4mus/fedialgo
* demo app: https://github.com/michelcrypt4d4mus/fedialgo_demo_app_foryoufeed
@glyph i think the right UX probably involves having like 4 or 5 preconfigured settings ("prefer recent", "prefer friends", "prefer trending" - things along those lines) with an "advanced" interface for people who want to nerd out on 20 sliders.
@glyph i ended up adding a set of preconfigured settings to FediAlgo after this convo
video: https://universeodon.com/@cryptadamist/114395249311910522
⚯ Michel de Cryptadamus ⚯ (@[email protected])
Attached: 1 video Ω🪬Ω New release of #FediAlgo (customizable #algorithm for your #Mastodon timeline) has a couple of cool features: 1. Configuration presets (so you can easily put discussions or trending toots at the top of your #timeline without fiddling with the individual settings) 2. A "What's Trending" section that will show you the top trending hashtags, links, and posts scraped from 30 or so of the most popular Mastodon servers All the old features like filtering for particular languages / hashtags / users or a minimum number of replies / boosts / etc. are still there. For now you still need to use the command line to run it (it's easy tho). If there's interest I can deploy the demo publicly. * Usable demo: https://fedialgo.thms.uk/ * Code: https://github.com/michelcrypt4d4mus/fedialgo_demo_app_foryoufeed * Library: https://github.com/michelcrypt4d4mus/fedialgo #FediTools #FediTips #Fediverse #Mastodon #activitypub #mastohelp #Fedi #foss #nodejs #opensource #MastoAdmin #Feed #timeline #algorithmicFeed #algorithmicTimeline #TL #algorithm #node #nodejs
Jeff Miller (orange hatband)