The Shifting Dynamics of Online Knowledge Platforms and the Implications for Generative AI Sustainability

Abstract

Generative Artificial Intelligence (GenAI) tools have fundamentally reshaped how users seek and share knowledge on digital platforms, raising critical concerns about the sustainability of usergenerated content (UGC)-the primary resource for training and advancing future AI models. While studies document an overall decline in UGC following GenAI adoption, understanding the heterogeneous effects across different user groups remains crucial for predicting the future of human-AI knowledge sharing. This study examines the impact of GenAI, specifically ChatGPT-3.5 release, on user interactions across Stack Exchange platforms, with a focus on Stack Overflow. Using a natural experiment approach, we analyze changes in activity patterns and interaction dynamics across user expertise levels, measured through platform reputation scores. Our findings reveal a striking pattern: while GenAI adoption has substantially reduced participation from low-reputation users, likely decreasing repetitive and basic questions, high-reputation users show no significant decrease in their overall activity. Instead, these experienced users have shifted toward increased peer-to-peer interactions, suggesting a transformation in how expert knowledge is shared and maintained. The results indicate that GenAI serves as a replacement for basic interactions among low-reputation users while potentially enhancing collaboration among high-reputation users, fostering a complementary relationship between AI and human expertise. This dual role suggests a possible pathway for sustaining high-quality UGC essential for training future AI models, even as overall platform activity declines.

Keywords: GenAI, Knowledge-Sharing Communities, Stack Overflow, ChatGPT