Member-only story
Maybe TikTok’s Algorithm Isn’t Particularly Remarkable, After All
It may be just bog-standard recommendation tech — and the real magic is in the UI
Like many people, I’ve read tons of pieces about the genius of TikTok’s recommender algorithm.
It’s certainly true that the system is awfully good. I’m a guitarist, and TikTok very rapidly grasped that fact — such that within a few days of signing up, my “for you” feed was full of kickass teens shredding Slayer and dobro-slide octogenarians belting out Delta blues. It’s pretty inspirational stuff! Many people I know tell me they have the same experience: TikTok’s algorithm just “got” them, quickly.
The feed is not all sweetness and light, mind you. Depending on your demographic, TikTok’s algorithm can lead you to some rather dark places. In a recent study by the Center for Countering Digital Hate, researchers found that when they created new TikTok accounts claiming to be 13 years old, it was only a matter of minutes before they were getting recommended anything from tummy-tuck surgeries to pro-thinness videos and suicide-related material.
Whether one enjoys the recommendations or is alarmed by them, the assumption of many people is: “Damn, TikTok must possess some utterly cutting-edge algorithm that’s lightyears ahead of its competition.” Write-ups of TikTok typically refer to its recommendation system as “scarily accurate”; it works so “eerily well”, as others note, that it is “a reflection of my subconscious thoughts”. Surely there must be some skull-king algorithmic magic at work here, yes?
Maybe not. I recently read a counterargument which suggests the opposite. Maybe the algorithm isn’t any great shakes; in fact, it’s probably quite mundane. What makes TikTok’s recommendations perform so eerily is not, in fact, rooted in the ruleset of its algorithm at all.
The success lies elsewhere: It’s in TikTok’s ruthlessly clever design, and its elegant UI.
This is the contention of Arvind Narayanan, a computer scientist at Princeton University, in his post about “TikTok’s Secret Sauce”.
Narayana starts by pointing out that TikTok’s algorithm is, by all accounts, a pretty bog-standard “recommendation system”…
