My research focuses on understanding the interaction between humans and advanced information technologies, with a particular emphasis on how users engage with algorithms and artificial intelligence (AI) in digital platforms, especially two-sided platforms.
I am interested in the societal and economic impact of these technologies. Specifically, I study how platform design, algorithmic interventions, monetization strategies, and policy decisions influence the behavior and outcomes of users such as content creators and viewers, as well as the overall performance of the platforms. Methodologically, I collaborate with leading digital platforms and use field experiments, quasi-experimental designs, and econometric modeling to identify causal mechanisms.
My work is anchored in two main research streams:
(1) the societal and economic impact of digital platforms;
(2) the societal and economic impact of AI.
Join work with Ao Huang, Nina Huang, and Renyu Zhang
Major Revision, MIS Quarterly (Job Market Paper)
Abstract: Digital platforms are increasingly deploying competitive features to drive user engagement and monetization. This work investigates the impact of social comparison, in the form of player knockout (PK) events, on streamer performance in livestreaming platforms. The PK events create a public, competitive situation wherein streamers perform acts in real-time and compete for viewer support through virtual gifting. Meanwhile, streamers face a central trade-off on whether to join PK: potential risks like negative self-evaluation, public humiliation, and viewer disillusionment versus possible benefits like increased visibility, fans, and earnings. Analyzing granular data from a leading livestreaming platform, we report two empirical studies. In Study 1, we employ Double Machine Learning (DML) and find that, on average, streamers participating in PK events increase total gifting by over 300%, viewer engagement duration by over 70%, and significantly boost streamers' follower growth. In Study 2, we leverage the platform's randomly matched PK events to analyze how competition structure affects streamer outcomes. Interestingly, our results show that when facing with bigger opponents (streamers with relatively more followers), although streamers are more likely to 'lose' in PK, they still benefit substantially from increased gifting and follower growth. Specifically, competing against either bigger opponents or those in the same content category increases both total gifting and cross-streamer follower attraction for the focal streamer. Our further heterogeneity analysis reveals two optimal PK configurations: small streamers benefit most from competing against bigger, same-category opponents, while big streamers gain most from competing against bigger peers in different categories, in terms of gifting and follower dynamics. Our study adds to the social comparison theory by examining social comparison in a public, real-time, and multi-party setting, and extends the social exchange theory by conceptualizing and measuring two distinct types of social exchanges, namely, reputational social exchange (reflected by cross-streamer follower migration) alongside traditional monetary social exchange (reflected by gifting). Our findings also provide actionable insights for platforms operators and streamers on designing and refining competitive strategies for enhancing user engagement and monetization.
Join work with Nina Huang and Renyu Zhang
Major Revision, Information Systems Research
Abstract: To maximize viewer engagement, online short-video-sharing platforms like TikTok often prioritize the established, star creators in viewer traffic allocation. However, small creators, who have a smaller number of followers but constitute the majority of the creator pool, are vital to these platforms, bringing diverse content and fresh perspectives. In collaboration with a leading online short video sharing platform, we leverage two large-scale randomized field experiments to evaluate the value of small creators to viewer engagement and examine the efficacy of viewer traffic allocation on small creator development. In the viewer-side experiment, 50% of the videos from small creators were removed from the recommendation algorithm's candidate pool for treatment viewers, but the control viewers received no intervention. We find that, on average, reducing the supply of small creators' content benefited viewer engagement on the platform. Furthermore, such treatment increased video watching time for less active viewers but decreased it for active viewers. In the creator-side experiment, the viewer traffic for treatment creators was boosted via recommendation algorithms, but not for control creators. Compared to the control creators, the treated creators produced 5.87% more videos without compromising content quality, exerted more effort in video production and self-engagement with their content, and enjoyed a 289.81% increase in the new follower count growth. The effect of additional viewer traffic was more pronounced for more experienced, popular small creators. Altogether, our research offers actionable insights into viewer traffic allocation that balances viewer engagement and creator development on short-video platforms.
Join work with Renyu Zhang
Join work with Xinyu Xu, Nina Huang and Renyu Zhang
Join work with Xinyu Xu and Renyu Zhang