Exploring Creative-Cultural Bias in the Design of AI Music : A Critical Analysis of Three Generative AI Music Services

Sammanfattning

This thesis examines the nature of design-induced bias in generative AI Music tools. Creative-AI, defined here as systems in which human creators share agency with machine learning models, is increasingly prominent in music production. A growing number of services, including the studied tools of AIVA, Boomy, and Mubert, allow users to generate musical recordings by manipulating only a limited set of parameters. These systems are often framed as democratizing creativity, yet they are embedded with cultural, technological, and market-driven assumptions that warrant critical analysis. Building on Friedman and Nissenbaum’s concept of algorithmic bias, Lindgren’s theory of AI assemblage, and Born’s dimensions of musical diversity, the thesis adopts a social constructionist perspective to situate music technologies within their cultural and material contexts. Methodologically, the study combines qualitative content analyses of tool interfaces with semistructured interviews with developers. These data were interpreted through thematic analysis, allowing for an exploration of how musical parameters, design decisions, and commercial priorities interact. The findings highlight three main concerns. First, music is predominantly conceptualized in terms of Western popular traditions, reduced to abstractions such as genres or moods, and with further parameters like harmony or tempo. This narrow framing limits expressive possibilities and obscures alternative musical ontologies. Second, the tools emphasize user agency, personalization, and accessibility while concealing aesthetic and technical constraints behind opaque interfaces. Users are encouraged to trust the system’s abstractions rather than engage critically with them. Third, the tools are closely aligned with platformcapitalist logics, encouraging users to participate in entrepreneurial practices and the monetization of creative labor. Rather than enforcing explicit musical biases, the systems appear shaped by the imperatives of online content commodification and engagement. By unpacking these dynamics, the thesis contributes to the emerging field of AI Music studies. It argues for more critical, socially situated approaches to the design and evaluation of AI Music tools, emphasizing that bias is not merely a technical flaw but an inherent feature of these assemblages.