Wits School of Arts (ETDs)

Permanent URI for this communityhttps://hdl.handle.net/10539/37947

Browse

Filters

20241

Settings

Author: search.filters.author.Armstrong, Douglas ConnollyStart date: 1995

Search Results

Now showing 1 - 1 of 1
  • Thumbnail Image
    Item
    MUCUS (Music Composition User System): Infectious Flexible Creative Interaction with an Algorithmic Music Composing Application
    (University of the Witwatersrand, Johannesburg, 2024) Armstrong, Douglas Connolly; Crossley, Jonathan; Harris, Cameron
    Recent research in the field of computational creativity and interaction design suggests new ways in which computers can contribute to a creative process. Computational creativity has necessitated new rigour in theoretical definitions of creativity for computational applications. Interaction design has evolved from a focus on efficiency and productivity to a user-centred focus on the emotional and hedonistic aspects of interaction with a computer. I set out to design an interactive algorithmic music composing tool that implements co- creative strategies for a human–computer collaboration, as described by Kantosalo and Toivonen (2016). This interactive tool would play the role of a creative collaborator in the development of musical material by the use of rule-based algorithmic music composing models, steering the co-creative composing process using high-level musical descriptors and capturing user sentiment to build a model of the user’s musical preferences. It is suggested that, through this process, an engaging creative interaction between human and computer can be sustained. Three versions of the software were tested in three different settings. The results suggest that a variety of co-creative and creativity support approaches are required to cater for a computational agent that does not match a musically trained human’s ability to identify musical merit in a developing idea. In this respect, it was found that there is further scope for exposing details of the computer decision-making process for development alongside the process of musical development, as a black box process of computational reasoning was found to be mysterious and at times frustrating. A rule-based system of music generation was found to be effective in a steering mechanism that matched higher-level descriptors of the musical variation process to music generation parameters. Engagement was sustained for longer when the duration of the musical output was longer in form. This included the ability to integrate the output of the application with existing digital audio workstations.