“Improvising virtual dancer meets audience: Observations from a user study” presented by Berman and James

Keywords: Dance, Improvisation, Creativity, Art installation, Humancomputer interaction, Expressive movement, Machine learning.

The interdisciplinary project AI am explores the potential roles of artificial intelligence in dance. One of the aims of the project is to create a performance where a human and a virtual dancer interact in real time in a mutual exchange where the human dancer and the avatar learn from and inspire each other. Apart from enabling an engaging experience, this kind of interaction would potentially also encourage movements that might otherwise not emerge from human improvisations. As a first step towards this goal, a method for generating novel dance movements from recorded motion sequences has been created. The next step is presented in this paper: an interactive setup where visitors can control parameters governing the avatar’s improvisations. A user study of the installation indicates that the avatar’s movements are interesting, engaging and inspiring. Implications of these results for future work are also discussed.