News Diversity Under Algorithms: The Effects of Pre-Selected and Self-Selected Personalization on Chinese TikTok (Douyin)

Existing journalism scholarship has raised the concern of people's limited news exposure on algorithm-driven social media. Using a series of novel agent-based testing (ABT) experiments, this study collects data from Douyin (the Chinese version of TikTok) and attempts to capture the dynamics of news diversity with complex interaction of algorithm and human agency. Results indicate that the personalized news consumption on Douyin is more diverse in news categories than it would be when occurs by chance. Additionally, algorithmic decisions more than human choices promote exposure to diverse topics in news consumption. We have further quantified whether users can obtain a more balanced news consumption through modifying their interest scope and tendency to engage with unmatched content. These findings are crucial for us to grasp how algorithms shapes users' news consumption and determine the effectiveness human users can adjust their exposure to a more diverse, opinion-challenging news environment.