Uncategorized

Following the Not So Online

Nadezhda1906

Bryan inspired me today:

feed request: just my timeline, minus anybody who posted more than 12 times in the past 12 days

I’d been thinking along the same lines recently. I didn’t realize it at first, but this is a big part of how I curate my follows. Occasional life events and baby photos, sure. Flood of memes and politics dunks and inscrutable subtweets, not so much.

I like social media! But only to a point. I’m down for 15 or 30 minutes of social scrolling a day, which isn’t all that much time, all things considered, so every new follow needs to pull their weight. I love all my Very Online friends, but I can’t often justify following them and their floods of Very Online Poasts. It’s just too much. I wish I could still use social media to keep up with them, but oh well. Them’s the breaks, I guess.

…or are they? Could we train a model to extract the occasional life event or baby photo out of a Very Online flood? I’d pay good money for that.

In one sense, this seems like what algorithmic feeds are aimed at. In practice though, they often end up optimizing for engagement more than human “importance.” Not too surprising for the big ad-supported platforms; engagement is their moneymaker. Plus, if you need to come up with a training dataset, it’s way easier to quantify and measure engagement than importance. Human labelers may be able to do it, sometimes, but us humans don’t come cheap.

So here we are. Barring someone with deep pockets funding a big new dataset, I’m stuck following the Slow Poasters and the Only Kinda Online. Maybe that’s not so bad.

Standard

23 thoughts on “Following the Not So Online

  1. I feel like this is the balance most One True Feed attempt go for. I want to be self-aware and chose a gear ratio to match my mood though

  2. Mayyybeee, on their noblest days, in spirit. But in practice those one true feeds seem to optimize for engagement more than “importance,” right? Not to mention, if you’re training a model, engagement is way easier to generate label data for than importance.

  3. a related idea is multiple degrees of “follow”, like “follow up to 1 post a day” or “follow every single thing for this account” or “follow for karma but don’t actually include in regular timeline”

  4. i’m pretty cynical about what mega-platform feed design goals and assessment metrics are as well, I was thinking of more sincere attempts

  5. yes! lots of prior art on all this in feed readers and their wide range of UX experiments. inbox, river of news, equal real estate per site, magazine, slide show, etc.

  6. bring me your finest “human importance” meats and cheeses and algorithms pls

    (no but seriously)

  7. Speaking of feed readers and inboxes, could we have a notion of read/unread message in bsky or alternate client? It doesn’t seem like rocket science.

  8. @KhΓΌrt I can’t. πŸ˜€ I expect it varies from person to person.

    Here’s one heuristic that would work for me: a post is more important if it’s about the person who posted it, or at least related to them, more than the latest internet meme or debate or news .

  9. @Nish great call! FraidyCat is a really useful idea in this space. Great low tech technique with a ton of bang for the buck. Plus Kicks Condor hangs out with us all in the IndieWeb pretty regularly, they’re great.

  10. Last week, I updated my blogroll to include everyone in my RSS feed reader. While I read a lot of topical blogs and newsletters, I…

Leave a Reply

Your email address will not be published. Required fields are marked *