Galactic Companions from Star Wars: C-3PO or R2-D2? Understanding Dual Types of Anthropomorphism in Humanoids

초록

Recent advancements in generative AI and humanoid robotics have facilitated the rapid development of humanoids that think, move, and interact in ways that closely resemble human behavior, significantly contributing to the study of anthropomorphism in human-AI interaction. Despite the extensive research in this area, the dynamic nature of anthropomorphism—how it evolves through cognitive processes and the effects of these changes— remains largely unexplored. Therefore, we propose dual types of anthropomorphism: reflexive anthropomorphism (automatic attribution of humanlike characteristics) and reflective anthropomorphism (conscious attribution of humanlike qualities) based on the dual process theory of higher cognition. A preliminary study was conducted with 120 participants using a scenario survey that included images of humanoids with and without human-like faces, as well as a video of a simulated chat interface between the user and a chatbot. The study result highlights that an AI agent's appearance significantly influences both reflexive and reflective anthropomorphism, with reflective anthropomorphism being particularly important for shaping users' anthropomorphic responses.