From perceptual rule transformation to listener attribution judgments: a blind-listening experiment on AI-generated, human–AI collaborative, and human-composed music

Human–AI emotional resonance in generative music: a mini-review of authorship attribution, affective response, and aesthetic evaluation

Generative artificial intelligence is increasingly involved in music creation, raising important questions about how listeners emotionally respond to music that is generated by, or attributed to, AI. This mini-review synthesizes recent empirical research on listeners' emotional, aesthetic, and physiological responses to AI-generated music. Existing evidence suggests that AI systems can reproduce some acoustic cues associated with basic emotional categories and may elicit measurable affective or attentional responses under specific conditions. However, listeners' evaluations are shaped not only by acoustic features but also by authorship attribution. Several studies report lower liking, perceived quality, authenticity, or engagement when music is attributed to AI, although null and occasionally reversed label effects have also been observed. To explain this pattern, this review proposes a dual-process framework that distinguishes between low-level acoustic–emotional processing and high-level source-based appraisal involving agency, intentionality, authenticity, and human experience. Overall, current evidence supports a partial dissociation between immediate affective responses and higher-order cognitive evaluation. Whether AI-generated music can support deeper forms of emotional resonance, such as long-term attachment, perceived understanding, personal meaning, or source-based empathy, remains insufficiently established. Future research should use identical-audio attribution designs, stricter audio-quality matching, cross-genre comparisons, and longitudinal listening paradigms.

From prompting to professional judgment: Generative AI, metalinguistic competence, and human-centered continuous improvement in advertising agencies

This study examines how generative artificial intelligence reshapes creative work in advertising agencies and considers its implications for professional capability development and continuous improvement. Attention centers on everyday routines, redistribution of expertise and organizational conditions that convert AI-enabled iteration into learning or, conversely, efficiency without capability growth. An exploratory qualitative design draws on open-ended questionnaire responses from 18 advertising professionals, including copywriters, art directors and video designers/makers. Researcher-led thematic analysis was supported by InfraNodus Lab. Text-network outputs served as sensitizing maps of recurrent concepts and semantic connections; final themes resulted from repeated comparison with complete responses. Four themes organize reported experience: generative AI as everyday creative practice, reconfiguration of creative process, time compression and process optimization and skill reconfiguration with deskilling risk. Participants associated AI with ideation, drafting, visual exploration and alternative generation. Their accounts also relocated professional value toward prompting, selection, evaluation, refinement and strategic interpretation. Metalinguistic competence appears as a capability for translating strategic intent into machine-readable instructions. Advertising agencies should integrate generative AI through reflective routines that preserve human judgment, professional learning and junior skill development. Training should combine prompt literacy with brand interpretation, output evaluation and shared review practices, while performance systems should assess capability growth alongside speed and productivity. Context-specific evidence from advertising agencies clarifies how metalinguistic competence links AI-mediated creation with professional judgment. A PDSA-informed model distinguishes reflective AI use, where generated alternatives are studied and converted into organizational learning, from transactional use, where output speed may rise without sustained skill development. Contribution rests on specifying a quality-management mechanism through which AI-supported iteration can become human-centered continuous improvement.

Understanding Listener Perceptions of AI and Human-Composed Music in Emotional Applications

Authorship cues and aesthetic judgement in the age of AI: Evaluating AI-composed music and human-composed music among higher education music students.

The present study investigates how the type of composer and the disclosure of authorship jointly influence aesthetic evaluations of music, with perceived authorship attribution examined as a mediating cognitive mechanism. Using an explanatory sequential mixed-methods design, the study involved music undergraduates in two experiments and follow-up interviews. Experiment 1 tested blind-listening evaluations and found no significant differences between AI-composed and human-composed music in perceived creativity and expressiveness. When authorship was disclosed in Experiment 2, AI-composed music was evaluated as significantly less creative and expressive, whereas evaluations of human-composed music were largely unchanged. Mediation analyses further showed that these effects were mediated by perceived authorship attribution. Notably, this mediating mechanism emerged only in the AI-composed music condition, indicating an asymmetric attribution process. The interview findings supported and contextualized the quantitative results by showing how listeners reinterpreted authorship meaning when evaluating AI-composed music. Together, these findings suggest that aesthetic judgements of AI-composed music are shaped less by perceptual qualities than by identity-driven cognitive attribution, thereby extending current understanding of authorship effects in musical aesthetics.

Comparing human and nonhuman judgments of perceived similarity of gender diverse talkers

Listeners conduct fine-grained analyses of gender from speech, and their perceptual organization of voices is presumed to be grounded in structured acoustic-phonetic information. Here, we compare human perceived similarity to automated self-supervised speech representations from a deep-learning model (i.e., HuBERT), which provides pairwise distance scores that increase with divergence in acoustic representations. Comparing human similarity judgments with HuBERT-derived distances tests the extent to which listeners’ perceptual organization is driven by acoustic representations versus socially- or task-driven constraints. Stimuli consisted of one sentence produced by 20 talkers representing five gender identities (cisgender man, cisgender woman, transgender man, transgender woman, nonbinary). In a free classification task, listeners (29 cisgender and 29 gender diverse) first grouped talkers by general similarity and then by perceived gender identity. Across all listeners and tasks, higher perceived similarity was associated with smaller HuBERT distances. This relationship was stronger in the unconstrained than the constrained task. Reliable alignment between human and nonhuman similarity metrics suggests that holistic acoustic distance partially drives how listeners classify talkers. However, human listeners also likely draw on knowledge of social constructs that automated similarity algorithms fail to capture, especially when the social category is invoked by task instructions.

Evaluation of Visual Metaphors of School Coexistence Using Generative Artificial Intelligence and Expert Judgment