题　目：Can Banning ChatGPT Save User-Generated Q&A Platforms?

嘉　宾：郑晋扬  助理教授  普渡大学

主持人：宋婷婷  副教授  上海交通大学安泰经济与管理学院

时　间：2023年07月12日（周三） 15:00-16:30

地　点：安泰经济与管理学院B201

内容简介:

The disruptions caused by ChatGPT have the potential to trigger moderating actions, including banning policies, for various affected business units. Among these units, the disruption and subsequent moderation policies on user-generated Q&A have far-reaching implications. User-generated Q&A serves as a crucial learning source for large language models (LLMs), making any qualitative and quantitative changes to user-generated Q&A capable of influencing the future development of LLMs. This study empirically examines the impact of banning ChatGPT-generated content on online Q&A platforms, focusing specifically on Stack Exchange forums. By analyzing the natural experiment of ChatGPT banning on select Stack Exchange forums and utilizing a generalized synthetic control method to address empirical challenges, the study sheds light on the effects of such a ban. The findings reveal negative quantitative effects and partially negative qualitative effects resulting from the ban. Specifically, the number of answers provided on the treated forums diminishes, with no significant change in quality but an increase in question length. Moreover, there is a decrease in the quantity of questions, particularly affecting high-quality and solved questions. While the average question quality remains unchanged, these questions attract less viewership and tend to be longer in length. Mechanism analysis further indicates that the decrease in the number of answers acts as a mediator for the declining trend in the number of solved questions. These findings highlight the adverse impact of banning ChatGPT-generated content, which contradicts the intended goal of maintaining a vibrant and useful Q&A platform. Consequently, they underscore the importance of adopting a balanced approach that addresses any negative aspects associated with AI-generated content, while ensuring the continued development of valuable Q&A platforms for users.

演讲人简介：

Jinyang Zheng is an assistant professor of management at Purdue University's Daniels School, specializing in information systems. With a focus on an economic perspective and predominantly empirical approaches, his research examines four main streams: managing content platforms, moderating online two-sided platforms, human-AI interactions, and digital resilience, employing diverse methodologies, including causal inference techniques such as Rubin's and Pearl's frameworks, along with structural models and unstructured data analysis. His research findings have been published in esteemed business journals such as Management Science, Information Systems Research, and MIS Quarterly, and have been recognized through numerous accolades, including the 2023 Jay Ross Young Faculty Scholar Award at Daniels School and multiple best paper awards or nominations at premier conferences, including ICIS, WISE, CIST, and CSWIM.

Jinyang completed his Ph.D. degree in information systems from the University of Washington in 2017. Prior to that, he obtained his bachelor's degree from the Department of Statistics at Fudan University in 2013.

欢迎广大师生参加！