嘉　宾：Miao Yu 博士生候选人 香港大学

主持人：房思含 助理教授 上海交通大学安泰经济与管理学院

时　 间：2025年12月3日（周三）13:30-15:00

地　 点：上海交通大学徐汇校区安泰楼A305

内容简介:

As food delivery platforms face growing pressure to balance operational efficiency and societal responsibility, their algorithmic systems increasingly balance between business goals and worker welfare. This paper investigates how algorithmic design has evolved from merely optimizing prediction accuracy to addressing broader organizational and welfare goals. Specifically, we study conservative Estimated Time of Arrival (ETA) algorithms that intentionally sacrifice some accuracy to alleviate rider stress and enhance societal welfare. We analyze how these conservative algorithms influence rider behaviors and earnings, examining whether riders simply adhere to adjusted time constraints or actively adapt their behavior to maximize income.

Using a mixed-methods approach that includes a field experiment and a sharp regression discontinuity design (RDD), we examine the impact of conservative algorithms on rider and consumer behavior. The field experiment reveals that highly conservative ETAs (e.g., a five-minute extension) increase riders' earnings through strategic order batching while simultaneously reducing total working hours, without negatively affecting consumer evaluations.  Our RDD, leveraging a natural 30-minute ETA threshold, demonstrates that shorter ETA extensions decrease delivery delays, improve per-order income, and enhance consumer satisfaction.  A survey of active food delivery riders further shows that in labor-abundant markets like China, income maximization strongly drives rider behavior, explaining their proactive adjustments to delivery strategies in response to algorithmic changes. Further heterogeneity analysis indicates that experienced riders benefit more from conservative ETA algorithms. Theoretically, our findings offer causal insights into how algorithmic designs shape gig workers' behavior. This study also positions ETA algorithm design as a valuable approach to mitigating the autonomy paradox and embedding ethical considerations within platform governance. Practically, our results suggest algorithmic adjustments can be an effective strategy to simultaneously enhance platform efficiency and worker welfare in the gig economy.

演讲人简介：

Miao Yu is a PhD Candidate in Innovation and Information Management at HKU Business School, The University of Hong Kong. Her research examines how algorithmic systems shape decision outcomes in high-stakes environments. Using large-scale administrative and platform data, she studies how digital policies and algorithmic designs influence user behavior, operational efficiency, and service quality. She also develops large language model based systems for public-sector and financial applications, including retrieval-augmented generation. In addition to her academic research, she has industry experience as an applied scientist at Ping An Technology and as a machine learning scientist at Meituan. She holds an MS in Technology and Operations from the University of Michigan and a BS in Management Science and Engineering with highest honor from Nankai University.

欢迎广大师生参加！