Music of 18 Performances: Evaluating Apps and Agents with Free Improvisation

Abstract

We present a study where a small group of experienced iPad musicians evaluated a system of three musical touch-screen apps and two server-based agents over 18 controlled improvisations. The performers’ perspectives were recorded through surveys, interviews, and interaction data. Our agent classifies the touch gestures of the performers and identifies new sections in the improvisations while a control agent returns similar messages sourced from a statistical model. The three touch-screen apps respond according to design paradigms of reward, support, and disruption. In this study of an ongoing musical practice, significant effects were observed due to the apps’ interfaces and how they respond to agent interactions. The “reward” app received the highest ratings. The results were used to iterate the app designs for later performances.