On recommendation algorithms
Inspired by @elomatreb's post, https://social.elomatreb.eu/objects/c5dbddf0-73bc-4196-9c62-ef922c03fb92 .
I've actually gotten a lot of mileage out of YouTube recs, and I respect the idea. But I am also very careful with what I allow the site to know I've watched. And I think underneath all of these machines is the incorrect assumption that interests can be deduced from just browsing habits. If I stopped watching a video partway, I could have wanted to finish it later, left out of disgust, or simply gotten bored. If I finished the whole thing, I could have really liked it, but I could've disliked it and was waiting for a payoff, or I could've even left it playing accidentally. I would love it if Twitter let me query for accounts followed the most by people I follow, but instead I get recommendations based on who it thinks I know IRL (I don't), what profiles I visited last (probably so I could mute them), or advertiser-friendly pigeonholes called "topics." Even Pandora, which I still hail as a gold standard for this sort of thing, partly because it actually lets you give explicit and non-side-affecting feedback, still doesn't know when I'm looking for something new or just the regular; something exciting or something mellow. I guess it would be nice if we could start working our way towards a middle between search and discovery; more exposed query functionality, more explicit control over rec engines.
The social network of the future: No ads, no corporate surveillance, ethical design, and decentralization! Own your data with Mastodon!