Why unequal AI access enhances team productivity: the mediating role of interaction processes and cognitive diversity

Abstract

Introduction: Generative artificial intelligence (GenAI) is widely viewed as valuable for improving the performance of human-agent teams (HATs). However, in reality, not all members have equal access to AI tools, making uneven AI integration an important factor impacting team composition and, thus, team effectiveness. While unequal access might seem detrimental, potentially hindering technology utilization, it could also foster deeper interactions and diverse expertise. To clarify these mechanisms, this study extends the classic Input-Mediator-Output model to an Input-Process-State-Output (IPSO) framework. Methods: A lab experiment involving 60 two-person teams was conducted, with teams assigned to unequal, full, or no AI access conditions. Results: The findings indicate that unequal AI access yields the highest productivity, improving both task quality and completion time compared to no or full AI access. This effect is driven by two key mechanisms. First, negative socio-emotional interactions and increased cognitive diversity serve as a positive serial mediation pathway linking unequal AI access to enhanced task quality. Second, unequal AI access leads to more concentrated and imbalanced questioning behaviors, which accelerates task completion. Discussion: This study provides an in-depth theoretical explanation of how AI integration structures operate in HATs and offers a foundation for strategically optimizing GenAI access in human-agent teaming.