Quadruped Robots and Canine Likeness: The Uncanny Valley Effect

This paper presents canine likeness features in robotic quadrupeds that influence their social perception. We adopted Contrastive Language-Image Pre-Training (CLIP), a neural network that has demonstrated signatures of the Uncanny Valley effect, to explore how the perception of quadrupeds evolves as their level of canine likeness intensifies. Seven models were tested, ranging from a fully robotic quadruped to a living dog with 252 images. Our findings indicate that the Uncanny Valley effect also develops in quadruped robots. This finding is a reference to selecting an appropriate level of realism for canine likeness fourlegged robots in Human-Robot Interaction (HRI).