Algorithmic Aesthetics into the Course:An Analysis of the Path of Post-humanistic Art Education Based on the Empirical Research of Traditional Music

This study presents a comprehensive network meta-analysis (NMA) of 16 empirical studies comparing listener evaluations of human-performed traditional music to three types of AI-generated renditions (generic AI, fine-tuned AI, and hybrid AI-human). Following PRISMA-NMA guidelines, we synthesized Hedges' g effect sizes for aesthetic ratings (e.g., beauty, authenticity, and likeability) across these conditions. Network geometry reveals a fully connected loop (Human–Generic–Fine–Hybrid). Results show that human performances were rated significantly more favorably than purely generic AI renditions (g≈+0.60, 95% CI [0.35, 0.85]), with moderate advantages over hybrid (g≈+0.20) and smaller differences versus fine-tuned AI (g≈+0.10). Fine-tuned AI slightly outperformed generic AI (g≈–0.40), and hybrid AI (combining AI and human elements) also modestly exceeded generic AI (g≈–0.30). These effect sizes (small to medium, according to Cohen's conventions) suggest that anthropomorphic elements (animacy, human-like "fit") drive preference, aligning with research showing that human-like AI voices boost likability. We discuss the implications for science and STEAM education: algorithmic music can serve as a vehicle for AI literacy and computational creativity in curricula, but educators must address concerns about perceived authenticity and fairness. Drawing on insights into anthropomorphism and Cheng's recommendations for AI in music education, we argue that science educators should incorporate AI-generated music thoughtfully—using it to engage students in computational thinking, ethics, and digital literacy, while also considering the equity of access. The paper concludes with future directions on ethical design, inclusivity, and the evolving role of AI in musical learning and science pedagogy.