Exploring the Behavior of Users “Training” Douyin’s Personalized Recommendation Algorithm System in China

This study investigates a novel form of human-algorithm interaction where users “train” TikTok’s recommendation algorithm. Through observations and interviews with 22 self-reported algorithm trainers, researchers identified the motivations, methods, and perceptual influencing factors underlying this behavior. The findings reveal that users employ various strategies, such as adjusting the functions of the recommendation algorithm system. Motivated by the pursuit of values and risk avoidance, this behavior is influenced by users’ algorithmic literacy, satisfaction with recommendations, and perceived benefits and costs. The research significantly contributes to our understanding of how individuals positively perceive and engage with algorithmic systems, shedding light on previously overlooked aspects. It introduces a novel strategy on addressing algorithmic risks from the user’s standpoint by advocating cooperation with algorithms.