Find the Good. Seek the Unity: A Hidden Markov Model of Human-AI Delegation Dynamics
研究管理者与AI系统间的委托动态，通过875名店长的纵向数据发现，管理者对AI能力的认知会导致委托意愿两极分化，形成反馈循环，高委托意愿者绩效更优。
As AI becomes integral to enterprise decision-making, this study explores the collaborative dynamics between managers and AI systems, focusing on human willingness to delegate tasks to AI. Grounded in the “agentic” systems delegation framework and instance-based learning theory, we employed a hidden Markov model in a longitudinal study of the dynamic delegation decision-making process involving 875 store managers. We found that there is a potential polarization in managers’ delegation willingness, with managers who recognize the capability of AI exhibiting high delegation willingness and fostering increased collaboration with AI over time—in contrast to their counterparts who are inclined to reduce AI’s involvement. During human-AI interactions, managers’ continuous performance appraisal of AI shapes their dynamic delegation willingness, which in turn affects their assessment of AI capability. This process forms a delegation feedback loop that drives the dynamics of delegation behaviors. Our study indicates that managers with a high willingness to delegate tend to outperform their counterparts and offers valuable insights for human-AI collaborative intelligence in organizational settings.