Dancing Through the Uncanny Valley: On the Likeability of Model-Generated Dance Movements

Abstract

The esthetic perception of model-generated dance movements was analyzed by varying the human likeness of the observed movements through the implementation of a model based on movement primitives. Likeability ratings and electrodermal activity in response to these model-generated movements varying in the number of underlying primitives were acquired for dancers and dance-novices. The results show that movements with a high degree of human likeness were generally associated with higher esthetic valuations. In dance experts, however, results show an uncanny valley effect in that likeability dropped for the most human-like model-generated kinematics. Additionally, the motion energy of the movement sequences also shows a remarkable association with esthetic responses, although to a lower degree for dancers than novices. Overall, these results not only extend the perceptual phenomenon known as the uncanny valley to the esthetic perception of model-generated movement kinematics, but also emphasize the relevance of sensorimotor experience for the esthetic perception of expressive movements.