Can AI-generated music induce emotions? A mixed-methods research study on the emotional impact of AI-generated music

Generative AI and its capabilities are impacting the traditional notions of music production, as the continuous development of music generation systems grants a fundamentally different perspective on music production and the way humans engage with music. While there indeed is potential in this development, AI-generated music is perceived by many as incapable of having any aesthetic impact as it would be unable to evoke emotions, which is the primary reason humans listen to music. Suppose Generative AI is to be of any significance, in that case, it is crucial to understand how humans engage with AI-generated music by studying the emotions experienced towards this music and if any intent is ascribed to it, two essential factors in the aesthetic experience of music. Unfortunately, this aspect has been underexposed in scientific research and requires additional attention. Therefore, this thesis investigates how listeners perceive the emotional impact of AI-generated music. This is done using a mixed-methods research strategy, combining an experimental- and phenomenological approach. A single-factor within-subjects experiment was designed to present musical stimuli of both human- and AI origin, assessing listeners’ emotional responses and perceptions of intentionality. This data was further explored through semi-structured interviews to investigate the phenomenon thoroughly. The thesis findings reveal that listeners experience emotions towards AI-generated music and generally ascribe intent behind the music, even when perceiving a piece of music to be of AI origin. However, a vital factor is how this music is presented, as the context in which the music is played, and the listener’s perception of the artist are crucial factors in the acceptance of AI-generated music. While the thesis is primarily exploratory and should be considered within the context of a Master’s thesis, it provides unique contributions to human-computer interaction. It challenges the traditional notions about AI-generated music, suggesting it cannot be emotionally impactful. The thesis findings give new insights into how this music could be applied in society and provide a fundamental stepping stone for further research.