1. Balancing Efficiency and Consumer Experience: An Investigation of Gig vs. Traditional Customer Support. (Job Market Paper)
Co-authors: Xiang Hui, Fuqiang Zhang, Tianjun Feng (academic); Yitong Wang, Congyi Zhou (industry)
Status: Under Review at Management Science; SSRN
Selected Talks & Conferences: INFORMS 2025 (Job Market Showcase), DSI 2025 (Doctoral Research Showcase), BU Online Research Seminar on Digital Businesses (2025), CIST 2025.
Award: Finalist, Best Presentation Competition (Doctoral Students), 2025 Fisher AI in Business Conference
I designed and executed a large-scale randomized field experiment at a leading e-commerce platform, assigning 265,967 consumer inquiries to gig (algorithm-managed) or traditional (human-managed) agents, and combined payroll data, consumer sentiment analysis, and machine learning classification of chat transcripts. The study shows that gig agents excel on algorithm-monitored efficiency metrics (4% higher closure rates, 35% shorter service durations, and 30% quicker responses) but generate worse consumer experiences, with 7% more negative emotions and 1% fewer follow-up orders. A key mechanism is strategic gaming: nearly half of gig agents’ case transfers are unnecessary escalations designed to maximize pay but that ultimately erode consumer trust. This paper provides the first causal evidence of the efficiency-experience trade-off in algorithmic management and highlights how narrowly designed metrics can create incentive misalignments.
2. Supply Elasticity at Algorithmic Pay Cliffs: Evidence from Gig Service Agents
Co-authors: Xiang Hui, Fuqiang Zhang, Tianjun Feng (academic); Congyi Zhou (industry)
This project studies how algorithmic and AI-based compensation systems influence labor supply and effort in platform service operations, reflecting my broader research on Empirical Operations Management and Artificial Intelligence Applications. Using proprietary data from thousands of gig agents, we exploit algorithmic “pay cliffs”, daily discontinuities in quality-based rankings that create quasi-random changes in pay, to identify causal labor-supply responses through a regression discontinuity design. The analysis reveals substantial and heterogeneous elasticity: lower-ranked agents work longer hours, exert greater effort, and achieve higher consumer satisfaction when near thresholds, while top performers show limited responsiveness. These incentive effects persist over time, suggesting adaptive expectations and strategic behavior under algorithmic management. Ongoing analyses model agent decision-making and evaluate alternative, fairness-aware incentive designs. Together, this research provides one of the first causal examinations of algorithmic compensation in practice and informs how AI-managed systems can balance efficiency, equity, and worker well-being.
3. Contract Design and Generative AI in E-Commerce Dispute Resolution
Co-authors: Fuqiang Zhang (academic); Congyi Zhou (industry)
Conferences: Presented at POMS 2025.
This project develops a principal-agent framework to study how generative AI reshapes incentive contracts in dispute resolution. I model AI copilots as probabilistic signals that alter the effort-accuracy trade-off, and I am implementing a mixed factorial experiment (AI availability × case complexity × AI accuracy) using state-of-art generative AI suggestions to test the model from pilot tests to field experiments. Current theoretical results show that higher AI accuracy reduces the need for intensive monitoring and strong performance bonuses, shifting contracts from complementary bonus-penalty schemes to substitutable ones, with effects most pronounced when baseline human judgment is unreliable. The model also predicts non-monotonic welfare effects: while moderate AI reduces agent rents, highly accurate AI benefits both agents and platforms. Online experiments with MTurk participants and field trials with real agents are designed to validate the predictions, enabling structural estimation of effort costs and AI trust weights for contract calibration. This work provides both theoretical and empirical evidence for how generative AI reshapes contract design in moral hazard settings and offers guidance for building effective human-AI governance systems in service operations.
1. Qiaowen Guo, Fuqiang Zhang, Tianjun Feng. "The Hidden Costs of Emotional Labor: Unpaid Effort and Attrition in Call Centers."
2. Qiaowen Guo, Daniels Kaitlin, Panos Kouvelis. "Service Segmentation for On-Demand Platforms."