Behavioral Responses to Robust vs. Stochastic Optimization Recommendations in Inventory Decisions

This poster investigates how human decision-makers respond to algorithmic recommendations in inventory management under uncertainty. While robust optimization (RO) and stochastic optimization (SO) models can significantly improve inventory outcomes, it remains unclear how managers perceive and adopt such advice. To address this gap, we design an online experiment using oTree, in which participants play an “ice bag ordering game” with uncertain weekly demand. Treatments vary across three conditions: no recommendation, RO-based recommendation, and SO-based recommendation. We also test how different data windows (12 vs. 60 weeks of historical demand) and cost structures (high-margin, low-margin, high penalty) affect behavior. The study provides behavioral insights into when and why people accept or reject algorithmic advice, offering practical guidance for designing decision-support tools managers actually use. The results contribute to advancing the field of Behavioral Operations Management by bridging optimization models and human judgment.