4 ms·

I just set up a Twitter account for the first time, to see what the Musk era is like. One immediate issue I notice is that the recommendation algorithm immedia

I just set up a Twitter account for the first time, to see what the Musk era is like. One immediate issue I notice is that the recommendation algorithm immediately starts siloing you into a roomfull of mirrors. Based on who you follow, it recommends similar people to follow (based apparently on who your followees follow plus maybe other data), so you quickly end up in an echo chamber. That's not the whole story, Twitter also tries to push-recommend 'noted celebrities' to follow, who can be summed up as the type of people who make the rounds of corporate media weekend talk shows. There doesn't seem to be any way to opt out of these recommendations. So far it's not a very illuminating experience. I notice that many people get around the character limit by posting images of text, which is an interesting work-around. Really, I think you'd need multiple Twitter accounts to get around this siloing effect. For example, on political issues, you could set up, say. 'socialist-communist-left', 'libertarian-capitalist', 'state-corporate-authoritarian', and 'alt-right MAGAs', 'politicians and media talking heads', etc. in order to get representative samples of the various competing US political views and actors. All in all I think it's kind of a waste of time, there are better information sources. [edit as others note, no real changes from pre-Musk era have taken place, so I'll just look at it once a week for a while]

jen20 4y ago
> Based on who you follow, it recommends similar people to follow I would actually be rather happy if this were the case beyond the trivial examples. For example, if I were to create an account and follow someone from the Linux kernel or open-source licensing space, and I got recommendations of other people working in that are, that would be _great_. Instead, I got Joe Rogan and AOC forced into my timeline because that is what maximizes rage and thus engagement. I'm out.
- mnw21cam 4y ago
  It seems like a hard problem to solve: I want to be siloed into subject interest groups, like to do with the various hobbies and work that I'm interested in. But I don't want to be shoved into an echo-chamber for politics. How does the algorithm distinguish between subject interest and politics? The difference is that for subject interest I'm looking for information specific to my subject, but the algorithm optimises instead for groups that agree with my opinion. I'd be quite happy to get information about my subject interest even if I don't agree with it. Then also the algorithm optimises for high drama, but that's a separate problem.
passwordoops 4y ago
What you describe is actually the "pre-Musk" Twitter - I created a new account late last year and it was as you describe. It has been an excellent source of information for me, but I am sure to follow good accounts (authors, podcasters, and their guests if they are interesting and don't RT junk), and sticking with timeline view. Unfortunately that's not how the majority use Twitter, so in the end your take is the correct one for the 99%...
rejectfinite 4y ago
>to see what the Musk era is like. You think they have had time to change anything yet?? jesus As someone who has worked for companies being bought, changes takes TIME.

jen20 4y ago

> Based on who you follow, it recommends similar people to follow I would actually be rather happy if this were the case beyond the trivial examples. For example, if I were to create an account and follow someone from the Linux kernel or open-source licensing space, and I got recommendations of other people working in that are, that would be _great_. Instead, I got Joe Rogan and AOC forced into my timeline because that is what maximizes rage and thus engagement. I'm out.

mnw21cam 4y ago
It seems like a hard problem to solve: I want to be siloed into subject interest groups, like to do with the various hobbies and work that I'm interested in. But I don't want to be shoved into an echo-chamber for politics. How does the algorithm distinguish between subject interest and politics? The difference is that for subject interest I'm looking for information specific to my subject, but the algorithm optimises instead for groups that agree with my opinion. I'd be quite happy to get information about my subject interest even if I don't agree with it. Then also the algorithm optimises for high drama, but that's a separate problem.

passwordoops 4y ago

What you describe is actually the "pre-Musk" Twitter - I created a new account late last year and it was as you describe. It has been an excellent source of information for me, but I am sure to follow good accounts (authors, podcasters, and their guests if they are interesting and don't RT junk), and sticking with timeline view. Unfortunately that's not how the majority use Twitter, so in the end your take is the correct one for the 99%...

rejectfinite 4y ago

>to see what the Musk era is like. You think they have had time to change anything yet?? jesus As someone who has worked for companies being bought, changes takes TIME.