Computational Model of Joint Human-AI Decision Adaptation

A novel computational framework explores the optimal distribution of decision-making responsibilities between human and artificial agents, emphasizing collaborative sequential adaptation. This model categorizes agents based on their memory types: humans rely on recency-weighted memory that favors recent results, whereas AI employs uniform memory that treats past outcomes equally. Rooted in the NK/NKC model lineage, it manipulates task scope (N), within-task coupling (K), and inter-agent coupling (C) across both modular and sequential designs. Significant discoveries reveal threshold dynamics that lead to stable high- and low-payoff states, prompting adaptation to inherited conditions. This research offers insights into effective task distribution within hybrid human-AI systems.