The Psychology of AI-Generated Music: How We Perceive and Respond to Machine-Made Melodies
There is an ongoing debate around the ethics of AI-generated music, with many arguing it takes away from the essence of human creativity.
But can our brains tell the difference between AI versus human-composed music, and can AI music evoke the same emotions as that created by humans?
The use of AI has taken over various fields, including creative ones such as music, poetry and art. AI-generated music is becoming increasingly popular, with various tools on the rise which compose anything from melodies to full pieces of music, using algorithms and machine learning. Often, these pieces of music are indistinguishable from those composed by real musicians. So, how do we as humans evaluate, perceive and accept such AI-generated musical works?
AI music generators
AI-generated music is revolutionising the music industry, creating an ethical divide between artists and music lovers alike. If you aren’t familiar with the capabilities of AI music generators, take a second to visit Udio. Click play on any piece you like. Now, tell me — how do you feel about it? And more importantly, if you didn’t know it was AI-generated, would you be able to tell? As a musician myself, this is both a scary and impressive realm to explore. What does the future look like, for both artists and audiences? I don’t think we can answer that for now, but what we can do is explore how we currently perceive and respond to AI-generated music, and how this may impact the future role and impact of such outputs.
Perception and composer bias
Perception plays a crucial role in how we respond to music. Likely, if an artist you aren’t so fond of releases a catchy song, we find ourselves irritated if we enjoy it, and often we may not even give songs by this artist a chance. Similarly, research has shown that when listeners know a piece of music is AI-generated, they like it less¹. This seems to be based on a pre-held bias, where scepticism around AI-generated music lies within its ability to create authentic and emotional art forms comparable to that of a human artist or musician.
This has significant real-world implications. If people prefer music less simply because they know it was created by an AI, it indicates a reluctance to accept AI-composed music and enjoy it. Hearing music without knowing its composer can mitigate this bias, similar to how blind orchestra auditions allow judges to appreciate the music’s quality without considering the musician’s gender, as knowing the identity of the composer ahead of time sets this bias in place¹.
Human or not?
A similar study by Hong et al. (2022) revealed that those who perceived the AI music generator as a musician appreciated its songs more than those who did not². For those who accepted the AI music generator as a musician, its composition was seen as fulfilling an attributed role responsibility, which resulted in a positive attitude towards AI-composed music. On the other hand, those who did not recognise the AI music generator as a musician showed doubts about the legitimacy of AI-composed music and therefore do not appreciate its existence or value².
Higher ratings of anthropomorphism — the attribution of human characteristics — led to the AI generator being accepted as a musician. This is based on the idea that we feel more comfortable with things that look like us, as it helps us to make sense of unfamiliar agents and reduce uncertainty³. Fink (2012) found that anthropomorphic designs that imitate human appearance, behaviour and interactions contribute to robot acceptance and can elicit social responses, which has shown to be useful in cases such as music-making⁴. However, further research is needed to explore the impact of anthropomorphic designs specifically for AI music generators, uncovering their role in the perception, acceptance and emotional responsiveness to their compositions.
The creativity conundrum
Can robots be creative without humans? This is an idea that is debated by many, surrounding the idea of creative autonomy, as a machine’s creativity is often attributed to its programmer⁵. This is central to the ongoing ethical and legal debates faced with such tools by the music industry, around intellectual properties and copyright. Who should hold the copyright to an AI-generated song — the generator itself, or the human programmer?
In the early 1980s, David Cope began developing “EMI”: Experiments in Musical Intelligence, which he spent 30 years teaching how to generate music in the style of composers such as Bach. The revelation was met with anger and pushback from audiences, critics and record labels. After this, Cope created a successor called Emily Howell, focusing instead on composing novel works in a unique musical style. Some critics complained that Emily’s works, “though musically pleasing, were hollow, shallow, and lacking depth and heart”⁹. David Cope said:
Most musicians, academic or composers, have always held this idea that the creation of music is innately human, and somehow this computer program was a threat in some way to that unique human aspect of creation,” Cope said. “I have always refuted that by saying a human-built the machine, listens to the output, and chooses what’s the best. What’s less human about that than if I had taken years and years to just compose the whole thing myself?¹⁰
Despite the controversy, Cope’s work with AI-generated music continues to challenge traditional notions of creativity and authorship that we face today. His efforts demonstrate that AI can be a powerful tool in the hands of human creators, expanding the possibilities of musical composition, and the role of human and machine collaboration in the creation process.
Once trained with enough data, current AI music generators can compose pieces without any input at all, besides the type of music wanted as an output⁶. So in this case, who holds the creative autonomy? Interestingly, Jackson (2017) argues that there is no significant difference between humans and machines with creativity since both generate innovative ideas from old ones. However, this perspective evokes further debate.
Whilst AI can mimic existing patterns to create something new, human creativity is unique in the sense it is driven by emotional depth, personal experience and intention. Here, AI replicates the process of creativity, but the authenticity of human artistic expression remains distinct. I, for one, love reading websites like Genius, which dive into the creative processes and why behind my favourite songs and lyrics.
But in a world where AI can mimic such emotions so well that we can’t often tell the difference, we lose the stories and personal touch behind the art, which I think over anything, takes away from the experience. Not everyone cares about this aspect of art, though. If something sounds good or looks good, for some that is enough. And maybe sometimes, it is. But I hope we never live in a world where all art is generated, and at least hold on to the uniqueness of such processes.
Emotional response
AI is not perceived to feel emotion or have personal experiences, which may contribute to both its acceptance as an artist, as well as its ability to evoke emotion in us⁸. However, several studies have found that when unaware of the music’s origin, participants exhibited similar emotional responses to both AI-generated and human-composed music⁹.
Research seems to indicate that if the individual does not know the music’s origin, AI-generated compositions can evoke the same strong emotions in human listeners, despite lacking personal emotion. These findings challenge traditional views on the necessity of human emotion and experience in creating emotionally impactful music and art forms, which many find uncomfortable.
Overall, this suggests that aesthetic judgments are not a result of stimulus features alone, but contextual information about the composer of a piece of music can influence aesthetic judgments of that music¹.
Genre bias
A study by Shank et al. (2023) showed that participants were more likely to attribute an AI composer to electronic music as opposed to classical music, which they hypothesised to be due to the “AI-sounding” nature of electronic music, which is more computerised¹. People have pre-existing ideas of what AI music should sound like, and electronic music likely fits into their notions. This shows a systematic stereotype towards certain genres, perhaps attributed to their perceived complexity and skill required to compose them⁷.
To investigate this further in the same study, participants were presented with classical music excerpts only, but the composer’s identity was manipulated as either human or AI.
They found that those in the AI group rated the music as being significantly lower quality than those in the human or control group, which demonstrates the impact of composer bias over the genre itself. Considering both of these findings, Shank et al. propose that as AI-composed music becomes more acceptable and the genres and styles created become more diverse, these stereotypes and biases may diminish¹.
The takeaways
In exploring the impact of AI on creative fields such as music, poetry, and art, it is clear that AI-generated music, in particular, is reshaping industry norms and audience perceptions.
AI tools can now produce music that rivals compositions by real, human musicians, blurring the lines between ownership and creativity. The wider acceptance of AI-generated music relies heavily on perception and bias, where knowing a musical composition is AI-generated often biases listeners against it, based on pre-existing beliefs around autonomous creativity, emotional capability, and likely a fear of what we cannot yet fully understand.
However, studies have shown that when unaware of the composer’s identity, we can emotionally connect with AI-generated music in the same way as human-composed music.
This challenges traditional notions of artistic authenticity and raises ethical and legal questions about copyright and creative autonomy. As the capability of AI evolves, so too must our understanding of its role in artistic expression and the broader implications of its integration into creative processes.
Sources:
¹ AI composer bias: Listeners like music less when they think it was composed by an AI
³ On seeing human: a three-factor theory of anthropomorphism
⁴ Anthropomorphism and Human Likeness in the Design of Robots and Human-Robot Interaction
⁵ Art, Creativity, and the Potential of Artificial Intelligence
⁶ How AI Can Change/Improve/Influence Music Composition, Performance and Education: Three Case Studies
⁷ Instrumentational Complexity of Music Genres and Why Simplicity Sells
⁸ Dimensions of Mind Perception
⁹ Investigating Listener Bias Against Musical Metacreativity
¹⁰ Virtual composer makes beautiful music — and stirs controversy
