Promoting DEI via Operational Interventions

Joint work with Kamalini Ramdas and Monika Heller.
Major revision at Operations Research; available upon request.
Finalists for the 2025 POMS College of Behavior in OM Junior Scholar Paper Competition.
We ran online experiments to examine how prior ratings and instructor gender impact instructor evaluations, under identical teaching quality. Our experimental evidence reveals sizeable negative effects of publishing low prior ratings, concentrated among inexperienced learners and on instructor-focused outcomes. Also, both female and male learners exhibit bias against instructors of the opposite gender, mainly when they face a higher cognitive load due to time pressure. Furthermore, female instructors are more vulnerable to the "ratings trap" when starting with a low teaching rating.

Joint work with Kamalini Ramdas and Monika Heller.
We experiementally examine how different ratings display formats can reduce prior ratings bias while still providing clients with sufficient and useful information.
The typical ratings distributions are summmarized from an anonymized dataset of over 78,000 observations from a leading e-learning platform.
Experiments in progress. Financially supported by the Wheeler Institute for Business and Development.

Improving Public Services via Emerging Technologies

Joint work with S. Alex Yang and Chaoran Liu.
We empirically examine the impact of Generative AI on digital marketing content. While Gen AI enables small businesses to create marketing materials more efficiently, content produced by large language models may lead to greater homogenization, diminishing customer engagement and eroding brand distinctiveness over time. To assess this impact, we exploit the temporary ChatGPT ban in Italy during April 2023 as a natural experiment, focusing on its effects on social media marketing by small business owners.
Analysis in progress. Financially supported by the Institute of Entrepreneurship and Private Capital.

Joint work with S. Alex Yang and Yiangos Papanastasiou.
Generative AI enables search engines to deliver concise, summarized answers to users, enhancing convenience and efficiency. However, this practice may reduce website traffic for content creators, potentially undermining their incentives to produce high-quality content. We analytically examine the impact of generative search engines on online information ecosystems, and propose design strategies that balance user convenience with content creator sustainability.
Analysis in progress.

Joint work with Kamalini Ramdas and Monika Heller.
Generative AI democratizes access to knowledge but also poses risks, such as over-reliance and misinformation. We investigate how tuning Gen AI settings can effectively balance the precision and engagement in generated responses to address these challenges. Through both online and field experiments (e.g., LBS classes, Coursera courses), we aim to examine the impact of AI tool settings on individuals’ comprehension and decision-making.
Experiments under design. Financially supported by the Wheeler Institute for Business and Development.

Ph.D. candidacy paper. Joint work with Kamalini Ramdas and S. Alex Yang.
We examine how group testing can improve diagnostic testing. Group testing involves pooling and testing samples collectively, reducing the number of tests needed while maintaining accuracy. On the other hand, the effectiveness of group testing depends on the risk levels of individuals within the same group, which is private information. Our game theoretic analysis shows that assigning individuals with identical risk levels to the same group (i.e., assortative batching) minimizes test costs and maximizes social welfare. In addition, we propose a screening contract to reveal private risk levels by self-selection, demonstrating how health administrators can improve testing strategies through operational design. We conduct online surveys embedded with the design for empirical validation.
We plan to refine the model considering bounded rationality to align theoretical results and experimental evidence.

Master Thesis. Joint work with Jian Chen.
Air catering companies face challenges in managing unpredictable meal demands, as they must fulfill last-minute orders based on passenger counts with little time for emergency restocking. Prediction models are useful but cannot fully address the cost balancing between wasted meals and urgent restocking. We propose to set up an on-site warehouse at the airport that can quickly restock planes as needed. This warehouse would receive regular shipments from the factory, enabling faster response times and better inventory management. We analyze the multi-period inventory policy under this design and verify it with real-world data from Beijing Air Catering Co., Ltd.

Joint work with Xiaofang Wang and Guoming Lai.
For the products that provide not only intrinsic value from their functions but also stylish consumption experience, there often exist both Veblen and network effects. Some customers are more likely to purchase the product if fewer customers can afford it, while others might appreciate the existence of more peers. We study the market equilibrium under rational expectations to investigate the appropriate strategies for these products. The optimal pricing and quantity decisions reveal interesting insights about the effects of such mixed consumption externalities.
Published in Operations Research Letters, 2017, Vol. 45, Issue 61 .