Too Much of a Good Thing? Negative Impact of High-Performing AI Teammates on Human Performance

Abstract

AI agents are increasingly deployed as teammates to work with humans on creative tasks. These developments underscore the need to explore how human performance is affected by AI teammate performance. We investigate the effects of AI teammate performance on individual human team member performance, examining the underlying mechanisms and contextual factors moderating this relationship. Findings from three experiments reveal that AI teammate performance negatively affects team members’ performance influenced by trust in the AI teammate and social loafing tendencies. Specifically, high-performing AI teammates foster increased trust and social loafing tendencies in human team members, subsequently diminishing individual performance in brainstorming tasks. However, the negative influence of AI teammate performance on individual team performance is attenuated when the perceived performance of another human teammate is high. In addition, our findings show that the presence of a high-performing AI teammate reduces productivity disparities between high- and low-knowledge team members. This equalizing effect stems from the negative impact of high-performing AI teammates on the performance of high-knowledge members. Our findings contribute to the growing literature on the impact of the inclusion of AI teammates on human members’ performance and provide practical implications for integrating AI teammates in organizational settings, particularly for creative tasks.

More information about this seminar on the website of Vrije Universiteit Amsterdam.