How agent appearance and interaction mode influence user self-disclosure: mediating role of perceived novelty

As personalised services increasingly rely on conversational and embodied agents, exploring how agent appearance and interaction mode jointly influence user self-disclosure is important for informing system design. This study conducted a 2 × 2 between-subjects experiment to investigate the interaction effect of agent appearance (robot-like vs. human-like) and interaction mode (text-based vs. speech-based) on users’ disclosure willingness. We developed a fictitious tourism website featuring an intelligent agent. The agent requested personal information disclosure from the participants, and the participants could answer or decline to answer. In total, 64 participants were included in this study. Results showed a significant interaction effect between agent appearance and interaction mode on disclosure willingness. Speech-based interactions elicited higher disclosure willingness than text-based interactions when the agent looked human-like, whereas this effect was reversed when the agent looked robot-like. This interaction effect was mediated through perceived novelty. Specifically, speech-based interactions produced higher perceived novelty than text-based interactions especially for human-like agents, and perceived novelty positively affected disclosure willingness. These findings highlighted the importance of anthropomorphic congruence in the agent design features.