Research on the Differences in Emotional Expression Between Artificial Intelligence Music and Traditional Creative Music

Abstract

With the proliferation of generative artificial intelligence (AI) in the arts, AI music has entered mainstream aesthetic discourse and reshaped the creative ecology of the music industry. Prior research has predominantly focused on the technical aspects of AI music generation, while neglecting the effects of creator identity expectations on listeners’ affective evaluations and leaving a critical research gap. In this work, we examine whether creator identity labels drive differentiated affective evaluations of identical music, the dimensional heterogeneity of this effect, and the moderating role of listeners’ musical expertise, employing a mixed-methods approach combining literature review, a between-subjects controlled experiment, and qualitative analysis with a single-stimulus multi-label paradigm to strictly control music ontological variables. Results reveal that identity labels exert a statistically significant main effect on evaluations of emotional depth and authenticity (scores for the human-created condition are significantly higher), with no significant effect observed for the pleasure dimension, and musical expertise acts as a key moderator. This work advances understanding of the emotional nature of digital art, and provides empirical support for establishing standardized AI art evaluation frameworks and facilitating human-machine collaboration in music.