Improving Inventory Management Quality with Reinforcement Learning: AI versus Human Decision-Making
通过实验室和实地实验，研究了基于强化学习的库存决策工具在稳定环境下能降低库存成本，但在波动环境下表现有限，并探讨了管理者对AI技术的接受度。
SYNOPSIS A limited understanding of artificial intelligence (AI) behavior leads to the gap between academic research on developing inventory management models and practical applications. In this study, we employ reinforcement learning (RL)—a branch of AI that focuses on learning optimal actions through trial and error within an environment—to develop an artifact for inventory decision-making. Through a laboratory experiment involving supply chain professionals and a field experiment, our RL-based artifact demonstrates superior capability in reducing inventory costs under stable conditions, while exhibiting limitations in volatile ones. Practitioner interviews reveal both support and concerns about the artifact, offering insights into enhancing managerial adoption of RL-driven technologies. Whereas these findings are specific to the inventory application, the lessons learned can apply to other judgment and decision-making contexts. Overall, our findings underscore the potential of AI in improving management quality and provide new insights into the alignment between emerging technologies and management practices. Data Availability: Data are available from the authors upon request. JEL Classifications: C52; D25; M11; M41.