AI automatic decision in newsvendor model with Nash bargaining fairness concern

Abstract

This paper investigates the impact of artificial intelligence (AI) automatic ordering and producing decisions on fairness-concerned supply chains under the newsvendor model. We develop a dyadic supply chain model in which the manufacturer acts as the Stackelberg leader while the retailer serves as the follower in a push supply chain. In contrast, their roles are switched in a pull supply chain. We assume that only human decision-making leads to decision regret behavior, whereas AI-automated decision-making does not. Without adopting AI, our results show that fairness concern does not necessarily lead to a decreasing quantity in ordering or producing, which is different from most previous studies. Different from the prior findings, our work reveals that in binding equilibrium, if fairness concerns are considered, the order quantity will decrease, while in non-binding equilibrium, the order quantity may not necessarily be less than the previous results. Interestingly, when decision regret bias is considered for fairness-concerned decision-makers, we can obtain quantity coordination solutions for supply chains under specific conditions. With adopting AI, our results show that increasing fairness concerns are beneficial for improving the follower’s profit while at the expense of sacrificing the leader’s profit margins, while the leader can only benefit from AI adoption when the decision regret bias of the follower is relatively high. It is noteworthy that under certain conditions, AI automation may negatively impact the profits of both push and pull decentralized supply chains. For instance, in low-margin profit scenarios where decision-makers exhibit moderate regret bias and fairness concerns, such effects can emerge. This indicates that under specific circumstances, the human behavioral factors — regret bias and fairness concerns — may sometimes enhance the performance of decentralized supply chain members. Our research findings provide significant practical implications for the adoption of AI-automated decision-making in real-world supply chains.