People Do not Automatically Take the Level-1 Visual Perspective of Humanoid Robot Avatars

Taking the perspective of others is critical for both human–human and human–robot interactions. Previous studies using the dot perspective task have revealed that people could automatically process what other people can see. In this study, following the classical dot perspective task, we showed that Chinese participants could not automatically process humanoid robot avatars’ perspective when only judging from self-perspective (Experiment 1) or randomly judging between self and avatar’s perspectives (Experiment 2), and people’s anthropomorphism tendency was related to the efficiency but not the automaticity of perspective-taking. These results revealed that human–human and human–robot interactions might be different in the basic visual process, and suggested the anthropomorphism tendency in people as an influential factor in human–robot interaction.