Simpler is (Sometimes) Better: A Comparison of Cost Reducing Agent Architectures in a Simulated Behaviorally Driven Multi-Echelon Supply Chain
48 Pages Posted: 21 Oct 2022
Date Written: October 10, 2022
Operations Management endeavors to identify polices that improve real-world operating systems. However, these systems often consist of actors that do not behave fully rationally and to improve performance managers must predict how such boundedly-rational actors will behave. Approaches to this problem range from myopic, limited information decision rules to more modern, but data intensive machine learning methods. This work compares approaches to reducing costs in a simulated, behaviorally driven, multi-echelon supply chain by experimentally changing the features of polices, including complexity, adaptability, incentive structure, and information availability. Results show that that relatively simple ordering policies, especially when accounting for behavioral features of other agents, in these systems can have large cost reducing effects only marginally behind more complex methods. Additionally, for some conditions, greedy, myopic, and limited information decision rules can be cost reducing globally. Under plausible conditions, decision makers in supply chains with other behavioral actors need not be perfectly rational and can be locally focused while achieving global benefits.
Keywords: Behavioral Operations Management, Bullwhip, Dynamic Decision Making, Supply Chain & Inventory Management, Model Predictive Control, Deep Q-Network, Supply Chain Research
JEL Classification: M11,C44,C63,C45
Suggested Citation: Suggested Citation