I trained a simplified lookalike of TikTok’s algorithm on 1,100 users’ donated data, to model the “map” it uses to recommend content to users.

I led a collaboration across numerous desks in which we asked our readers to share their TikTok watch histories with us to help understand the algorithm. For this fun interactive, I trained a custom embedding model that translated into a (much requested by readers) map of TikTok explaining how the algorithm puts users into “niches”. The series won a National Press Club award, a SABEW award, and was a finalist for an Online Journalism Award. Methodology: https://www.washingtonpost.com/technology/2025/10/07/tiktok-app-study-user-behavior-addiction/