Find the Good. Seek the Unity: A Hidden Markov Model of Human-AI Delegation Dynamics

This paper applies the IS delegation framework to study store managers' willingness to delegate product replenishment decision tasks to Artificial Intelligence (AI). We develop a hidden Markov model (HMM) to longitudinally explore the delegation dynamics, where performance appraisals based on delegation outcomes may shape managers' future willingness to delegate. We find store managers' delegation decisions are influenced by perceptions of AI value in collaborative task performance based on their performance appraisals of delegation outcomes. Managers' confidence in AI delegation is built or eroded from repeated performance appraisals in a delegation feedback loop, which can polarize them into two extreme states of delegation willingness. Managers of high-delegation willingness display behaviors that are more adaptive when it comes to depending on AI for decision automation, whereas low-willingness managers are more stagnant. When delving into the delegation performance of a human-AI collaborative team, we see that high-willingness managers interact with the system more professionally and tend to achieve better team performance. Our findings offer useful insights on managing human-AI collaborative intelligence in the workplace.