PPIG 2018 - 29th Annual Workshop
Choosers: The design and evaluation of a visual algorithmic music composition language for non-programmers
Matt Bellingham, Simon Holland, Paul Mulholland
Abstract: Algorithmic music composition involves specifying music in such a way that it is non-deterministic on playback, leading to music which has the potential to be different each time it is played. Current systems for algorithmic music composition typically require the user to have considerable programming skill and may require formal knowledge of music. However, much of the potential user population are music producers and musicians (some professional, but many amateur) with little or no programming experience and few formal musical skills. To investigate how this gap between tools and potential users might be better bridged we designed Choosers, a prototype algorithmic programming system centred around a new abstraction (of the same name) designed to allow non-programmers access to algorithmic music composition methods. Choosers provides a graphical notation that allows structural elements of key importance in algorithmic composition (such as sequencing, choice, multi-choice, weighting, looping and nesting) to be foregrounded in the notation in a way that is accessible to non-programmers. In order to test design assumptions a Wizard of Oz study was conducted in which seven pairs of undergraduate Music Technology students used Choosers to carry out a range of rudimentary algorithmic composition tasks. Feedback was gathered using the Programming Walkthrough method. All users were familiar with Digital Audio Workstations, and as a result they came with some relevant understanding, but also with some expectations that were not appropriate for algorithmic music work. Users were able to successfully make use of the mechanisms for choice, multi-choice, looping, and weighting after a brief training period. The ‘stop’ behaviour was not so easily understood and required additional input before users fully grasped it. Some users wanted an easier way to override algorithmic choices. These findings have been used to further refine the design of Choosers.