报告题目：Writing Quality and Soft Information in the GenAI Age: Evidence from Online Credit Markets

报告时间：2024年10月15日19:00-20:00

报告地点：腾讯会议264567750

报告人：周文君 教授

邀请人：郭艳红 教授 

报告摘要：

  Writing matters in finance for both informational and non-informational reasons. Using data from a dominant online credit platform, we show that generative AI (GenAI) can significantly enhance the writing and perceived quality of loan applications, especially for high-credit borrowers because they treat good credit and quality writing as substitutes in applications, whereas they are complements in the data for receiving funding. However, our proprietary language models reveal that GenAI decreases soft information conveyed due to the convergence in writing, especially when borrowers deliberately prompt GenAI to improve funding probability rather than general writing quality. If lenders adjust their lending considering GenAI adoption, they recapture some soft information, mitigating potential credit misallocation. We also characterize the hypothetical long-run equilibrium in which GenAI adoption and responses in credit decision-making are rational and endogenized. Finally, we introduce a Writing Quality Index that is easy to compute and predicts heterogeneous fundability of borrowers across credit levels. The findings provide insights into the evolving role of soft information and potential impacts of GenAI in lending.

报告人简介：

Dr. Wenjun Zhou is the Lawson Professor of Business and serves as the Ph.D. in Management Science (Analytics) Program Director at the Haslam College of Business, the University of Tennessee Knoxville. Additionally, she holds the position of Chair of INFORMS College on Artificial Intelligence, a section of INFORMS (Institute for Operations Research and the Management Sciences). Her general research interests are data mining, business analytics, and statistical computing. She is passionate about developing methodology, models, and algorithms for discovering useful knowledge from data, which can support real-world decision making. Her work has been published in refereed journals and conference proceedings, including Machine Learning, INFORMS Journal on Computing, IEEE Transactions on Knowledge and Data Engineering (TKDE), European Journal of Operational Research (EJOR), ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD), and IEEE International Conference on Data Mining (ICDM).