Seeing eye to eye: trustworthy embodiment for task-based conversational agents.

Area of Science:

• Human-Computer Interaction
• Artificial Intelligence
• Robotics

Background:

• Smart speakers are common in homes but not workplaces due to privacy and trust concerns.
• Abstract conversational agents lack social cues, hindering workplace adoption.
• Embodiment may enhance trust and social presence for workplace AI.

Purpose of the Study:

• To investigate the impact of embodiment on trust, engagement, and performance in a workplace setting.
• To compare a humanoid robot, a smart speaker, and a dialogue manager in a first responder scenario.
• To assess the role of anthropomorphic features in user perception of AI agents.

Main Methods:

• Mixed-methods study involving a first responder team.
• Comparison of a humanoid robot, a smart speaker, and a custom dialogue manager.
• Evaluation of trust, engagement, cognitive load, and human performance.

Main Results:

• The humanoid robot, with facial features and eye gaze, was potentially more trusted than the smart speaker.
• Embodied conversational agents increased task engagement and performance compared to the smart speaker.
• Subjective trust is complex, but embodiment appears to positively influence user experience.

Conclusions:

• Embodiment, particularly with anthropomorphic features, may facilitate the integration of conversational agents into professional environments.
• Further in-situ studies are recommended to validate findings with domain workers.
• Robotic embodiment shows promise for improving human-AI collaboration in task-oriented settings.