Advertise here with Carbon Ads

This site is made possible by member support. โค๏ธ

Big thanks to Arcustech for hosting the site and offering amazing tech support.

When you buy through links on kottke.org, I may earn an affiliate commission. Thanks for supporting the site!

kottke.org. home of fine hypertext products since 1998.

๐Ÿ”  ๐Ÿ’€  ๐Ÿ“ธ  ๐Ÿ˜ญ  ๐Ÿ•ณ๏ธ  ๐Ÿค   ๐ŸŽฌ  ๐Ÿฅ”

How TikTok’s Algorithm Figures You Out

Using dozens of bot accounts, The Wall Street Journal did an investigation and determined that TikTok’s algorithm needs only one piece of information to determine what you want to watch: the amount of time you spend watching individual videos. Observing your watch time and rewatching is enough for them to fill your “For You” page with recommended videos that are right in your wheelhouse after just an hour or two. That this happens so quickly and completely โ€” 90-95% of what users see on TikTok is algorithmically determined โ€” leads to users going down narrow-interest rabbit holes that can be dangerous, e.g. if someone’s Covid interest turns into anti-vax QAnon crap or sadness turns into video after video about depression or harming yourself.

As someone who built an entire web app that collected people’s social media likes/faves, this focus on a single signal is fascinating. API limitations and rate limits on the number of requests would keep you from building a service with a TikTok-like algorithm for Twitter or Instagram that used likes as the only signal for whether to show someone a piece of content or not, but if you could, I bet it would be amazing.