Abstract

Machine listening and new musicology intersect in the contemporary study of genre detection, algorithmic bias, 
and canon formation, reshaping how music is classified, interpreted, and culturally valued. Machine listening 
employs computational techniques to extract musical information from audio recordings, while new musicology 
critically interrogates the sociohistorical, political, and institutional assumptions embedded within musical 
discourse. Together, these approaches illuminate how automated genre classification systems influence the 
organization and circulation of music in digital environments. This study examines the conceptual foundations of 
genre, the computational methods used in genre detection, and the evaluation metrics and datasets that underpin 
machine-listening systems. It further analyzes how sampling bias, algorithmic opacity, and institutional 
preferences reinforce unequal representations of musical traditions, particularly privileging Western popular and 
art-music canons over non-Western and marginalized genres. The study also explores how corporations, 
streaming platforms, and academic institutions shape contemporary canon formation through recommendation 
systems, metadata infrastructures, and large-scale digital archives. Through case studies involving commercial and 
academic datasets, the discussion demonstrates that computational systems are neither neutral nor purely 
objective, but are deeply influenced by historical, economic, and cultural assumptions. The paper argues that future 
computational musicology must prioritize transparency, reproducibility, inclusivity, and interdisciplinary 
collaboration in order to support more equitable representations of global musical cultures. Ultimately, machine 
listening and new musicology reveals both the possibilities and limitations of algorithmic approaches to music, 
highlighting the need for critical frameworks that balance technological innovation with cultural sensitivity and 
ethical responsibility.