AI composer bias: Listeners like music less when they think it was composed by an AI.

Abstract

The use of artificial intelligence (AI) to compose music is becoming mainstream. Yet, there is a concern that listeners may have biases against AIs. Here, we test the hypothesis that listeners will like music less if they think it was composed by an AI. In Study 1, participants listened to excerpts of electronic and classical music and rated how much they liked the excerpts and whether they thought they were composed by an AI or human. Participants were more likely to attribute an AI composer to electronic music and liked music less that they thought was composed by an AI. In Study 2, we directly manipulated composer identity by telling participants that the music they heard (electronic music) was composed by an AI or by a human, yet we found no effect of composer identity on liking. We hypothesized that this was due to the "AI-sounding" nature of electronic music. Therefore, in Study 3, we used a set of "human-sounding" classical music excerpts. Here, participants liked the music less when it was purportedly composed by an AI. We conclude with implications of the AI composer bias for understanding perception of AIs in arts and aesthetic processing theories more broadly. Public Significance StatementArtificial intelligence (AI)-computers making intelligent decisions or emulating humans-is revolutionizing the music industry. Yet, very little is known about how people emotionally respond to AIgenerated music. The findings of the present work indicate that listeners tend to be biased against music that they think was created by an AI if the music itself does not fit expectations of what an AI could create.

Context changes judgments of liking and predictability for melodies

“…Different liking ratings in the alone and combined contexts demonstrate that the same stimuli can elicit different responses in different circumstances. This is consistent with other research that indicates that the reward and pleasure associated with a stimulus are context dependent, namely relative to the set of alternatives in which the stimulus is embedded (Tremblay & Schultz, 1999) and the expectations of the individual (Belfi et al, 2021;Kolbeinsson et al, 2022;Shank et al, 2022). By changing the context in which our stimuli were presented, we aimed to induce a shift in the reference context used by participants for evaluating the stimuli.…”

“…Similarly, experienced reward has been shown to be influenced by comparisons to other rewarding stimuli in rats (Webber et al, 2016), and there is evidence that dopamine response to rewarding stimuli is sensitive to motivation and context (see Schultz, 2013 for a review). There is also behavioral evidence in humans indicating that context and expectations can affect aesthetic evaluation (Belfi et al, 2021;Kolbeinsson et al, 2022;Shank et al, 2022). Taken together, this evidence suggests that experienced pleasure for a stimulus is malleable and can vary relative to the broader reference context in which that stimulus is experienced.…”

Your ears don’t change what your eyes like: People can independently report the pleasure of music and images.

“…At first glance, our work also stands in contrast with other prior work which has found that context and extra-stimulus information does influence an observer’s aesthetic judgment of an object. For example, previous research has indicated that information about a musical performer (Shank et al, 2022) or title of an artistic work (Millis, 2001; Turpin et al, 2019) can have an influence on one’s aesthetic judgment of that work. However, in these cases, this extra-stimulus information can be presumed to be a part of the stimulus itself and to have originated from the same source (i.e., the artist).…”

Artificial Intelligence and Music. A Literature Review / Binomul inteligența artificială și muzica. Revizuirea literaturii de specialitate

“…An interesting experiment investigated the possible bias of individuals towards AI-composed music, showing no significant differences in participants' satisfaction after learning whether the music was composed by a human or AI (Zlatkov, Ens and Pasquier, 2023). Later research examined the same issue, however, finding a bias against AI-generated classical music, leading to a decrease in satisfaction with the music listening process (Shank Shank, Stefanik, Stuhlsatz, Kacirek and Belfi, 2023). In a study of 446 participants, both music lovers and music professionals, an overall negative perception of artificially composed music was observed (Tigre Moura and Maw, 2021).…”