We talked about the geopolitics of TikTok in TiB 124
and TiB 126
. Whether you’re a TikTok hawk or a TikTok dove, the key to understanding the app is the algorithm it uses to decide what to show you. Eugene Wei has a new-ish, brilliant post, “Seeing Like an Algorithm
” that walks through the design choices that make the algorithm work. It’s quite long, but worth it.
The core idea is what Wei calls “algorithm-friendly design”: the whole app experience is designed to give the model as much information as possible on what content you might like (and what you won’t). By contrast, existing social media giants like Facebook and Twitter don’t do this. Their “infinite scroll” interfaces optimise for lower scanning friction and positive-only interaction, which are key if you’re building a network of friends, rather than of interests.
This helps explain, among other things, why so many internet services offer such poor recommendations despite apparently knowing so much about us. As Wei says, it’s interesting to think about what a Twitter-like product built on “algorithm-friendly design” might look like and deliver. Most of the things I write about in Thoughts in Between I find on Twitter, but it’s extraordinary how much work building a finely tuned following graph requires. I expect there are lots of opportunities still to be uncovered in this space.