Mechanisms of Algorithmic Advice Dependence on Supply Chain Risk Cognitive Bias in Human AI Collaborative Decision Making: An Experimental and Structural Model Analysis

The integration of algorithmic decision support systems into modern supply chain management has fundamentally transformed traditional risk assessment practices. However, the underlying cognitive mechanisms through which human decision makers develop an over-reliance or dependence on algorithmic advice remain poorly understood in current literature. This study systematically investigates how algorithmic advice dependence influences supply chain risk cognitive bias within human-AI collaborative decision-making contexts. Particular attention is given to the New Zealand business environment, which currently faces significant geopolitical disruptions and severe commodity price volatility. Drawing extensively on behavioral decision theory and cognitive load theory, we propose a comprehensive moderated mediation model. In this framework, algorithmic advice dependence affects risk perception bias through the critical mediation of cognitive effort reduction, which is further moderated by task complexity and advice quality. Using advanced agent-based simulation experiments integrated with publicly available supply chain disruption databases-including the NZ Global Dairy Trade price index and geopolitical risk indices from 2020 to 2025-we systematically manipulate algorithmic advice characteristics to examine their direct effects on risk cognitive bias formation. Structural equation modeling is employed to rigorously test the hypothesized relationships. The results reveal that algorithmic advice dependence significantly amplifies risk perception bias, particularly under high task complexity conditions. Ultimately, this research contributes to the behavioral operations management literature by elucidating the cognitive mechanisms underlying human-AI interaction and offers practical implications for designing robust algorithmic decision support systems that effectively mitigate cognitive biases.