Lessons Learned from a Field Experiment on TikTok's Artificial Intelligence Systems and User Outcomes
Artificial intelligence systems, particularly in social media, are under increasing scrutiny (Reviglio & Agosti, 2020). TikTok’s recommendation system, renowned for its high degree of personalization, exemplifies this trend (Bhandari & Bimo, 2022). This lightning talk will present insights from an ongoing field experiment pre-registered involving 700 TikTok users. Participants were divided into two groups: a control group using TikTok as normal, and an experimental group that disabled personalization in their news feed for two weeks. Pre- and post-experiment measures include mental health, political polarization, and problematic TikTok use. By sharing our design, early findings, and challenges, we seek feedback from the scientific community to refine our approach. This talk also aims to spark discussion on innovative ways psychological researchers can study Human–AI Interaction, particularly in the context of highly adaptive systems like TikTok.