Songs have long served as an effective pedagogical medium, offering cognitive benefits such as facilitating memory and comprehension, as well as affective benefits such as lowering anxiety and enhancing motivation. Although K-pop has been widely used ...
Songs have long served as an effective pedagogical medium, offering cognitive benefits such as facilitating memory and comprehension, as well as affective benefits such as lowering anxiety and enhancing motivation. Although K-pop has been widely used for this purpose in Korean language education, its lyrics often contain colloquialisms, abbreviations, and fast tempos that limit classroom suitability, and instructors face difficulties in selecting pedagogically appropriate songs. Recent advances in generative artificial intelligence (AI) have made it possible for instructors without musical expertise to produce complete songs using text prompts alone. Despite this emerging possibility, no study in Korean language education has yet explored the direct development of educational songs through collaboration with generative AI. Addressing this gap, this study aims to develop educational songs for Korean language education through human-AI collaboration and to design instruction for their classroom use.
To this end, this study first analyzed the collaborative patterns between instructors and AI that emerged during the song development process. The results showed that lyric writing consistently followed a common collaborative pattern regardless of song type, whereas composition exhibited four distinct patterns—AI cover songs, AI arranged songs, AI co-composed songs, and AI-generated songs—depending on the degree of AI involvement. Based on these findings, this study derived a human-AI collaboration model and prompting guidelines, and developed five educational songs for Korean language education incorporating target vocabulary (derivatives, collocations, antonyms) and grammar items (numeral classifiers, reason expressions) using ChatGPT and Suno.
Prior to full-scale classroom application, a small-scale case study was conducted with 13 beginner and intermediate learners in a short-term Korean language program in Japan. Learner-centered activities were divided into conventional activities, such as lyric-based role play, and generative AI-based songwriting activities, in which learners composed their own Korean songs using generative AI. This pilot study identified necessary revisions when the developed songs were applied to a specific learner group and confirmed that gaps in learners' digital literacy influenced their satisfaction with AI-based activities.
Drawing on these findings, an instructional model was designed by applying a modified ASSURE model, incorporating AI literacy into learner
analysis and digital readiness into the utilization of media and materials. The resulting model integrates PPP-based, TTT-based, and song-based instructional approaches with the learner-centered songwriting activity.
This study makes the following contributions. First, it presents a concrete pathway for instructors to develop educational songs that overcome the limitations of existing songs such as K-pop, centering instruction on singing rather than mere listening. Second, it demonstrates the potential of AI-assisted songwriting as learner-centered instruction. Third, it establishes a systematic procedure linking song development and instructional design through a human-AI collaboration model and the modified ASSURE model. Fourth, it empirically identifies digital literacy gaps as a practical issue and reflects this finding in the instructional model. However, this study did not include quantitative verification of instructional effectiveness or expert validation of the developed songs, and further research addressing these limitations is needed.