Probabilistic Generative Music as a Data Sonification Medium

Auditory displays that incorporate musical structural elements and textures have been used to increase long-term listenability and engagement. Such approaches are often met with skepticism due to the inherent tradeoff between aesthetics and information fidelity. Additionally, their typical reliance on simple deterministic mappings (e.g. MIDI pitch or note duration changes) produce sonic outputs that are either overly predictable or musically incoherent. I propose and demonstrate a framework that expands the parameter mapping space through the manipulation of determinism and nondeterminism in real-time sequenced generative music. The result is dynamic and improvisatory musical patterns that may sustain listener interest while saliently conveying data trends with adequate perceptual independence between musical dimensions. The resulting parameter space may be suited to applications targeting artistic expression or augmented relaxation, sleep, and movement, where informational fidelity and temporal precision may marginally be sacrificed in favor of musicality and its associated affective / motivational benefits. Although pending formal evaluation, this work reflects flashes of the potential that probabilistic generative music has as a data communication medium.