Long-Term Interaction and Persistence of Engagement for Musical Interaction using a Genetic Algorithm

Abstract

Current research in human-agent interaction primarily focuses on short term interaction and rarely addresses day to day use. We propose a prototype system based on a genetic algorithm that places long term interaction as the core design goal. The goal of this system is to develop stand alone long-term development and provide a platform for future post-processing of deep learning generations. This paper addresses these issues through the domain of musical interaction and improvisation, a field that incorporates dialogue-like interaction built on stylistic constraints. We contend that the objectives of continual knowledge development and building relationships are key to long-term human interaction, and design the genetic algorithm specifically around these concepts. Our eventual goal of the prototype is a future application of post processing for deep learning generative systems.