Connected vehicles have become a promising platform for conversational agents. However, drivers might struggle to control systems that feature multiple AI agents. To improve the usability of increasingly complex systems and the user’s interaction experience, the integration of various conversational agents needs to be carefully considered. This study aims to investigate the following: What constitutes an efficient, user-friendly arrangement of the tasks performed by a home and a car artificial intelligence (AI) agent? What is the optimal method to trigger the use of such agents? A between-subjects factorial experiment was conducted with three types of AI agent setting (home AI generalist vs. vehicle AI generalist vs. home and vehicle AI specialists) and two activation methods (voice activation vs. push-to-talk activation). The results indicate that interacting with specialist AI agents enhances the perceived ease of use and the credibility of these agents. Furthermore, voice activation improves the social presence and attractiveness of an AI agent. The findings suggest that an AI agent that offers role specialization and a natural interaction method will improve drivers’ interaction experience.

The watching-eye effect proposes that others’ eyes cause people to behave in a prosocial manner. The current study tested this in the context of an online news website, by investigating whether a watching-eye icon influences users’ attention to themselves and expressions of their opinions in a comment section. In an online experiment, participants (N = 741) used an online news website in the presence (vs. absence) of a watching eye as a visual cue for an imagined audience, who reportedly presented their opinions in a comment section. Results showed that the watching eye did influence participants’ private and public self-awareness and the quality of their comments. Presence of the visual cues, compared to its absence, increased female participants’ self-awareness, specifically when others’ opinions revealed in the comment section were mixed or opposed to the news article topic. This increased private self-awareness was positively associated with the comment quality. These findings indicate the importance of social cues on interfaces in mitigating the negative consequences of anonymity in online environments.

The ambiguity of online text communication coming from the lack of audiences’ social cues often causes misunderstanding, which arouses a negative image of the message sender. Likewise, message senders feel social anxiety about whether their intentions are properly conveyed, or whether their messages may not sound offensive or impolite in communicating with text such as writing a self-introduction letter. In this paper, we design and develop a messaging assistant (Emoware) prototype which visualizes the audience’s predicted emotions to motivate the message sender to reveal more information. We conducted a user study comparing Emoware to plain text interface with 56 participants and found that Emoware positively affects writing by improving users’ self-disclosure intention, social support, and self-disclosure. This study contributes to overcoming the limitations of text modality in online communication.

The online environment for video conferencing lacks cues compared to offline, so one can hear the interlocutor's criticism more sensitively, and the fear of presenting in front of the camera can hinder participation in the meeting. It is known that interface design affords a role in improving public speaking and has a possibility of changing user behavior. To examine how the interface design of video conferencing affects video debating participation, 2 (visual anonymity: avatar vs. face) × 2 (self-affirmation vs. no self-affirmation) between-subjects experiment was conducted. Results showed that using an avatar, when properly used together with self-affirmation, has a positive effect on active participation in discussions, but derogating others’ critical messages. These results indicate unique underlying mechanisms of the effects of the avatar; the deindividuation effect of visual anonymity, and the effects of improving participation when customizing self-value reflected avatars.