Paper Type
Short
Paper Number
PACIS2026-1478
Description
This study examines how an augmentation-mode AI creation assistant reshapes creator performance on a user-generated content platform. Leveraging the introduction of a GenAI writing assistant on a leading Chinese shopping recommendation platform as a quasi-natural experiment, we apply machine learning and synthetic difference-in-differences to address three questions: who adopts such tools, how adoption affects creation quantity and quality, and how effects vary across creator capability levels. We find that adopters exhibit distinct content profiles, and that adoption simultaneously improves both creation quantity and quality—contrasting with the productivity–quality trade-off commonly observed under automation-mode AI. We further plan to investigate how these effects vary across creators with different capability levels to shed light on the boundary conditions of AI augmentation.
Recommended Citation
Wang, Siqi; Tong, Jack; zhang, xiaofei; and Lai, Kee-hung, "Adoption, Adjustment, and Achievement: Investigating How AI Creation Assistants Reshape Creator Performance" (2026). PACIS 2026 Proceedings. 5.
https://aisel.aisnet.org/pacis2026/ai_ml/ai_ml/5
Adoption, Adjustment, and Achievement: Investigating How AI Creation Assistants Reshape Creator Performance
This study examines how an augmentation-mode AI creation assistant reshapes creator performance on a user-generated content platform. Leveraging the introduction of a GenAI writing assistant on a leading Chinese shopping recommendation platform as a quasi-natural experiment, we apply machine learning and synthetic difference-in-differences to address three questions: who adopts such tools, how adoption affects creation quantity and quality, and how effects vary across creator capability levels. We find that adopters exhibit distinct content profiles, and that adoption simultaneously improves both creation quantity and quality—contrasting with the productivity–quality trade-off commonly observed under automation-mode AI. We further plan to investigate how these effects vary across creators with different capability levels to shed light on the boundary conditions of AI augmentation.
Comments
01-AIML