Understanding the processes of trust and distrust contagion in Human–AI Teams: A qualitative approach

Abstract

The success of human–AI teams (HATs) requires humans to work with AI teammates in trustful ways. However, trust does not exist in a vacuum but forms through and can be influenced by interactions among teammates, leading to understudied questions about how trust or distrust can be spread within a HAT. Drawing on interviews with 36 participants who collaborated in a three-member human–AI team, we explore human perceptions of and reactions to a human or AI teammate’s (dis)trust spread about an AI teammate, and uncover the process and impact of such spread. Our findings highlight that a trustworthy (dis)trust spreader can catalyze trust contagion within a human–AI team through various social and cognitive processes. We provide one of the first empirical investigations into specific ways through which trust or distrust can be spread within HATs and people’s perceptions of such spread. We thus contribute to the effective design of AI teammates and human–AI team dynamics that foster an appropriate level of trust in future HATs. • Trust contagion in HATs is driven by a trustworthy (dis)trust spreader via social processes. • Distrust contagion within a HAT can be driven by the spread of misinformation. • AI and human teammates must be “on the same page” and “on the same side” to build trust. • AI’s functional design can shape social perceptions and interactions.