Exploring the Interaction Between Anthropomorphic Appearance and Performance on Trust and Acceptance in Voice Assistants
Denise Sogemeier, Frederik Naujoks, Yannick Forster, Josef F. Krems, Andreas Keinath · International Journal of Human-Computer Interaction · 2025
In-vehicle voice assistants (VAs) can potentially enhance driving safety and user experience. However, their success depends on trust and acceptance. This study explores two influencing factors: (1) design-based and (2) performance-based attributes. In a driving simulator study, N = 53 participants engaged with VAs varying in anthropomorphic appearance and response latency. Results indicate that neither factor universally enhanced trust or acceptance. Instead, a nuanced interaction emerged: When response latency was short, anthropomorphic appearance had no impact. However, when response latency was long, both the non- and high-anthropomorphic VA were rated as more trustworthy than a low-anthropomorphic VA. Additionally, participants rated the high-anthropomorphic VA equally trustworthy across both latencies, suggesting a potential buffering effect of anthropomorphism under suboptimal performance conditions. Furthermore, trust fully mediated the relationship between the factors and acceptance, supporting models that conceptualize trust as a precursor of acceptance. Based on these insights, design recommendations were derived.