Intelligent voice assistants and chatbots feature a variety of anthropomorphic cues that can influence the perception and behavior of their users. For example, previous research showed that implementing faces in learning materials can increase learning performance. Furthermore, it was found that people would disclose more personal information if a chatbot featured a human name and face. It has been unclear to what extent the modality of intelligent voice assistants as an additional anthropomorphic cue compared to visual cues affects actual self-disclosure.
In an experimental study, participants interacted with either a chatbot or an intelligent voice assistant, both running on a smartphone. The conversational agent was presented either with the name "Jonas" and a smiling emoji as avatar or without a concrete name with an abstract avatar consisting of two circles. The conversational agent asked participants to answer personal questions of increasing sensitivity before and after the presentation of a learning module.
Our results show that participants revealed more private information after interacting with an anthropomorphic conversational agent than those who interacted with a neutral agent. No significant effect was found for modality on self-disclosure. In contrast to previous research, anthropomorphic cues did not foster learning outcomes. Our findings reinforce the promotion of privacy awareness for users of conversational agents.