Presentation
Less Trust, More Mirroring Words: Lexical Alignment as a Response to Uncertainty in Human-AI Teams
DescriptionUnderstanding trust contagion—how trust spreads among human operators toward an AI teammate—is critical for enhancing cooperation in human-AI teams. While previous research has examined trust contagion (Rojas & Li, 2024), the role of conversational alignment, specifically lexical and structural alignment, in facilitating this trust contagion process has received less attention. This study investigates how lexical and structural alignment contribute to this trust contagion process. By manipulating a confederate’s expressed trust level (high, low, neutral) toward an AI teammate, we analyzed how this influenced the linguistic alignment patterns of human teammates in collaborative interactions. Surprisingly, participants aligned their word choices (lexical alignment) more when their teammate expressed low trust, suggesting that linguistic mirroring may act as a compensatory behavior under uncertainty rather than a signal of trust. Structural (syntactic) alignment showed no significant differences across conditions. These findings reveal that people may unconsciously adapt their language more in low-trust environments to navigate uncertainty, rather than to build trust. Understanding these patterns can inform the design of conversational AI systems that monitor linguistic cues to assess and support team trust dynamics in real time.
Event Type
Lecture
TimeTuesday, October 14th1:30pm - 1:50pm CDT
LocationGrand B
Human AI Robot Teaming (AI)