entrain is a shared installation that uses active learning to stimulate social interaction in co-located mobile music-making.
Participants may fill circular sequences to generate rhythmic loops. Depending on their individual behaviour, the machine may designate specific participants by generating audiovisual feedback in an adaptive way. The resulting expressive workflow taking place between humans and the machine may foster social interaction through musical entrainment, that is, through rhythmic synchronization between humans and machines.
entrain was developed using a participatory design method. We started with an observation step to brainstorm interaction scenarios with stakeholders before deciding on the machine learning technique to be studied. This enabled us to identify active learning as a relevant technique for collective musical interaction. We then implemented a model prototype, called Bayesian Information Gain, which enabled to steer participants toward new musical configurations, while being sufficiently complex to appear as a black-box to them—which was of interest for such a public installation.
The project was developed with Abby Wanyu Liu, Benjamin Matuszewski, and Frédéric Bevilacqua in collaboration with the ISMM group of IRCAM, as well as Jean-Louis Fréchin and Uros Petrevski from Nodesign.net, and Norbert Schnell from Collaborative Mobile Music Lab of Furtwangen University, in the context of the Sorbonne Université Doctorate in Computer Science.
Installation at ACM SIGGRAPH 2019 Studio
Paper at ACM SIGGRAPH Studio (2019)
Available on GitHub