AI and music education

Recent studies (e.g., Chauhan, Reference CHAUHAN2017; Cheung & Slavin, Reference CHEUNG and SLAVIN2013; Delgado et al., Reference DELGADO, WARDLOW, MCKNIGHT and O’MALLEY2015; Kadiyala & Crynes, Reference KADIYALA and CRYNES2000) have highlighted the crucial role of technology in education, particularly its impact on enhancing student motivation and the overall effectiveness of both learning and teaching. During the COVID-19 pandemic, the shift towards technology was prominent, with extensive use of various formats, including virtual, synchronous, asynchronous, linear, web-based, distance, e-learning and blended learning, which have significantly advanced educational quality (Rapanta et al., Reference RAPANTA, BOTTURI, GOODYEAR, GUÀRDIA and KOOLE2021; Williamson, Reference WILLIAMSON2021). In addition, the rise of artificial intelligence (AI) in education marks a new era of exponential growth (Su & Yang, Reference SU and YANG2023). In December 2022, OpenAI garnered extensive media attention following the launch of a free preview of ChatGPT, an advanced AI chatbot powered by the GPT-3.5 architecture. According to OpenAI, the preview achieved significant engagement, with over one million user signups recorded within the first five days of its release (Roose, Reference ROOSE2022).

AI technologies are claimed to do more than merely respond to information queries; they anticipate needs, adapt to changes and proactively support students (Chen et al., Reference CHEN, XIE, ZOU and HWANG2020). This dynamic interaction between students and AI systems enhances adaptation and comprehension, allowing educational communities to evolve, reason and collaborate (Bajaj & Sharma, Reference BAJAJ and SHARMA2018). Research suggests that AI tools aid in decision-making, operate with a degree of autonomy and manage routine tasks, thereby freeing up time for other activities (Conklin & Hartman, Reference CONKLIN and HARTMAN2014; Chiu et al., Reference CHIU, MOORHOUSE, CHAI and ISMAILOV2023; Weatherly & Weatherly, Reference WEATHERLY and WEATHERLY2026). This shift contributes to enhanced learning experiences and provides crucial support in environments where traditional academic or social support systems may be lacking.

In music education, numerous high-quality programmes and apps are available for teaching various aspects of music, including theory, solfeggio, harmony, counterpoint and more (Zulić, Reference ZULIĆ2019). Examples include AIVA, Flow Machine, ORB Composer, Interval and Practica Musica. Some of these programmes aim to augment creative thinking and enhance students’ learning experiences. Wei et al. (Reference WEI, KARUPPIAH and PRATHIK2022) stressed that AI in music education, though still in its early stages, presents unique challenges and opportunities. For example, in Cooper’s (Reference COOPER2024) study, 56 assessors evaluated a total of 410 music lesson plans; each plan was given a quality score and labelled as either human-created or AI-generated. The results revealed that human-made lesson plans received higher quality ratings overall. However, assessors could not reliably distinguish between human and AI-generated plans, with an accuracy rate of only 55%. This discrepancy raises questions about the inherent value of human-created content compared to AI-generated content, challenging traditional views on creativity and expertise in lesson planning. Although music education teachers and students can potentially benefit from AI by it providing additional resources, enhancing practical skills and encouraging innovative thinking, its adoption lags behind other subjects due to economic, political and social factors (Zulić, Reference ZULIĆ2019). Whilst AI and music education are still at their early stages of development, many scholars (e.g., Holland, Reference HOLLAND and Miranda2000; Wei et al., Reference WEI, KARUPPIAH and PRATHIK2022) suggest that personalised music teaching approaches can support individual student needs, helping them achieve their full potential.

Theoretical framing: democratic pedagogy

Teachers play a crucial role in identifying each student’s strengths and systematically leveraging them to foster innovation. Interactive and personalised teaching methods can effectively integrate current educational theories at various stages of learning. Theoretically, I propose that advancements in AI-driven educational tools could offer a bridge between enhancing autonomy and personalisation in learning, potentially aligning with the deep-seated democratic values espoused by John Dewey. Dewey (Reference DEWEY1916) regarded democracy not merely as a governmental system but as a way of life, fundamentally linked to human relations and essential for personal development. He posited that for democracy, synonymous with education in his view, to thrive, educational environments must embody democratic principles. As Dewey highlighted, democracy in education is not just about administrative structures or policies but about fostering a culture that values individual development and communal interaction.

From this perspective, numerous scholars have explored how democratic principles manifest within educational contexts. Welzel (Reference WELZEL2002) observes that individuals with higher education levels and less traditional lifestyles tend to show stronger support for democratic ideals. The notion of a ‘democratic learning space’ (DLS) has been primarily theoretical in nature, yet Ade-Ojo and Duckworth (Reference ADE-OJO and DUCKWORTH2017) elaborate on this concept by describing DLS as environments where participation is equitable, and the authority of knowledge and its carriers is acknowledged. This framework is supported by theories that place DLS within broader ideological contexts. Furthermore, O’Quinn (Reference O’QUINN2005) highlights the importance of critical literacy skills, particularly the ability to decode socio-political statements and goals, urging learners to engage with these within their own cultural and historical contexts. Oni (Reference ONI2006) investigated how individuals’ perceptions of democracy are influenced by factors such as their educational background, cultural environment, living conditions and values. Complementing these views, Ade-Ojo and Duckworth (Reference ADE-OJO and DUCKWORTH2017) advocate for the creation of democratic learning spaces that empower learner agency, feature a participatory curriculum and employ pedagogical methods that support effective learning. From a humanistic standpoint, Veugelers (Reference VEUGELERS2011) views education as a moral endeavour aimed at supporting human development. He champions moral values as central to educational systems and individual teaching practices, advocating for reflective, dialogical and democratic learning as fundamental to a humanistic educational approach. These elements are deemed crucial for fulfilling humanistic educational goals. These insights collectively underscore the multifaceted relationship between democracy and education, emphasising the need for educational practices that foster both individual and collective growth.

In an extension of Dewey’s (Reference DEWEY1916) seminal exploration of democratic values in education, Bernstein’s work, Pedagogy, Symbolic Control, and Identity: Theory, Research, Critique (2000), introduces essential principles such as enhancement, inclusion and participation in knowledge production, thus crafting a robust framework for democracy within educational settings. Bernstein candidly acknowledges that these conditions for an effective democracy do not stem from ‘higher-order principles’ but rather from what he termed a ‘naïve condition for democracy’, which he then translated into the pedagogic democratic rights of enhancement, inclusion and participation. These rights are designed to operate across individual, social and political spheres, thereby embedding a deep structure of democratic engagement within the fabric of educational practice.

Within the music education context, Allsup (Reference ALLSUP2003) stressed that democratic education is more intricate and substantive than simply allowing choices on superficial matters; true democracy in education requires genuine collaboration and must engage not only adults but also incorporate the perspectives and rights of both teachers and students. However, arguably, traditional, authoritarian models of music learning are still more prevalent in formal educational settings, such as conservatories and schools, compared to more informal, participatory learning environments like teen rock bands or local choral societies. Silverman (Reference SILVERMAN2013) contends that democratic music education, rooted in care, reflection and respect, enables teachers and students to collaboratively deepen their musical understanding and appreciation of diverse identities. This shared process nurtures both personal and musical growth, cultivating empowerment, self-expression and a sense of belonging. Such an environment becomes transformative, fostering a collective metamorphosis within the educational experience. Moreover, from a pedagogical standpoint, since the 1980s, Swanwick’s CLASP model (Reference SWANWICK1979) has advocated for a holistic approach to music education, emphasising music approaches such as composition and performance in the classroom. When aligned with democratic learning principles, AI has the potential to serve as a transformative tool by expanding access to music-making, personalised learning experiences, such as leveraging technology in composition and performance and fostering critical engagement with these technological advancements.

Aims of the study

The theoretical framework of this study is deeply rooted in the exploration of the interplay between technology and pedagogy, particularly within music education. Guided by a desire to transcend mere empirical evaluation, this study seeks to explore the broader philosophical implications of integrating artificial intelligence (AI) tools within educational practices. This inquiry is centred around an in-depth, school-year-long case study of a singular music teacher and administrator at an international K-12 school in Macau, serving as a microcosm for larger educational enquiries.

I began by questioning whether GenAI could potentially align with democratic values by facilitating an educational experience that respects individual differences while promoting collective understanding and participation. This structural necessity demands the democratisation of educational settings that promote personalised learning. With the integration of GenAI, while teachers and students gain a higher degree of autonomy, particularly in facilitating more personalised learning, it also introduces significant ethical concerns in teaching and learning. Hesmondhalgh (Reference HESMONDHALGH, Curran and Hesmondhalgh2019) argued that uncritically praising the positive aspects of digitalisation tends to ignore its complexities and the broader economic and political forces involved. He contended that labelling these benefits as ‘democratic’ is overly simplistic and lacks nuance. Zulić (Reference ZULIĆ2019) critiques the paucity of scholarly inquiry into the application of generative AI in school-based music education. Kucirkova and Leaton Gray (Reference KUCIRKOVA and LEATON GRAY2023) emphasise that for AI to truly enhance its role in fostering democratic citizenship among children, its application in pedagogy must extend beyond merely tracking the technological aspects of personalised knowledge delivery. Instead, there should be a greater focus on the democratic implications of using data-driven systems in education, ensuring that these technologies support democratic values and contribute positively to the development of students’ identities as democratic citizens. Moreover, the integration of GenAI also brings forth challenges that must be navigated with a commitment to these democratic ideals.

In response to current gaps, this study explores the intersection of GenAI and democratic pedagogy within the contextual boundaries of an intrinsic case study. I posit that democratic teaching can manifest in any educational context, provided that it remains consciously upheld as an ideal by the educator. Furthermore, GenAI holds the theoretical potential to facilitate democratisation within the music classroom, fostering inclusivity and accessibility in music creation and pedagogy (Cheng, Reference CHENG2025). However, the extent to which this potential is actualised remains contingent on the critical engagement of the educators in shaping AI’s role within the learning process. Specifically, the study investigates the following:

RQ1: How do GenAI tools shape and influence the pedagogical strategies employed by the music educator?

RQ2: What are the main concerns and challenges when integrating GenAI tools into democratic music teaching practices?

GenAI in this context

I was inspired by the increasing popularity and usage of generative AI tools, particularly those based on large language models (LLMs), such as ChatGPT, Microsoft Copilot and Claude. These tools are referred to as the ‘GenAI tools’ in this study. These GenAI tools typically support users in personal assistance, educational support, content creation, programming help and creative writing via conversational AI. Currently, these GenAI tools incorporate the Generative Pre-trained Transformer (GPT) architecture, which uses self-attention mechanisms to determine the relevance of words in a sentence or context, enabling the model to understand relationships between words and their meanings. There are also other AI models highly relevant to music education, particularly AI composition tools. GenAI music creation tools, such as Suno AI, allow users to produce realistic songs by inputting text prompts, aiming to ‘democratise music creation’ (Cheng, Reference CHENG2025), making music creation accessible regardless of musical background. For instance, the Text-to-Music Generation function could enable students to compose music more easily and creatively.

In this study, Adam predominantly utilised LLM Gen AI tools such as ChatGPT and AI-assisted music tools like Soundtrap. Additionally, his students occasionally made use of Suno AI. All these tools Adam used offer free access for users. Though there are a few other paid AI music software available, their costs and availability greatly constrain access.

Triangulation

Numerous researchers have stressed the importance of employing verification strategies throughout the course of a research study in ‘constructing evidence within the qualitative project’ (Meadows & Morse, Reference MEADOWS, MORSE, Morse, Swanson and Kuzel2001, 187). This study utilised multiple methods that included ‘combining various analytic approaches, such as constant comparison, immersion/crystallisation, matrices, manual analysis and computer-assisted analysis, as well as employing two different methodological approaches to analyse the same data’ (Meadows & Morse, Reference MEADOWS, MORSE, Morse, Swanson and Kuzel2001, 194). To ensure robustness in our analysis, I initially coded the data from various sources. Subsequently, I enlisted a research assistant to verify the findings through manual thematic coding. This cross-checking process involved independently reviewing the data coded by me, followed by detailed discussions to resolve any discrepancies. This iterative approach ensured that our final interpretations were comprehensive and accurate. Through this collaborative process, we engaged in a detailed discussion to understand the nature of the differences and worked to reconcile them. This iterative process of comparison and re-analysis ensured that our final interpretations were robust and reflected a comprehensive understanding of the data.

Vignette

Adam (pseudonym) brings over a decade of teaching experience to his diverse roles in education, ranging from pre-kindergarten to college instruction. As a male in his mid-thirties, Adam’s keen interest in emerging technologies forms a core part of his daily routine. Trained initially as a band director, Adam has not only mastered multiple instruments but has also performed in a band with friends, where he often rotated between different brass instruments and occasionally played the synthesiser. Despite not considering himself an expert in music technology, Adam’s enthusiasm for the field is palpable. His engagement with technology took on new significance during the pandemic, which catalysed his integration of digital audio workstations into his classrooms. His aim is to bridge the gap between popular culture and traditional music education, making learning more relevant and engaging for his students.

At his school, Adam wears multiple hats. He teaches middle school music classes and plays a significant role in the curriculum team, overseeing the development and implementation of music curriculum and policies at the middle school level. His responsibilities are multifaceted: preparing students for the Cambridge IGCSE Music examination, leading a rock band after school and collaborating with other teachers on curriculum development. At the end of 2022, with GenAI tools becoming publicly available, Adam expressed a proactive interest in utilising GenAI such as ChatGPT and Copilot in his teachings. During our initial meeting, he shared that his school had begun to explore the potential integration of GenAI tools into the classroom setting.

Beginning in the fall of 2023, Adam’s school initiated a project through the curriculum team to delve into the integration of AI education within the classroom setting. Tasked with developing professional development programmes on AI for teachers, Adam embraced this opportunity to explore how GenAI tools could be effectively utilised within his own teaching context.

Setting

Adam is employed by a private international school in Macau. This K-12 school enrols around 700 students and teachers from over 20 countries. It operates under a conceptual framework that promotes core beliefs about the oneness of humanity, societal evolution, the purpose of education and the nature of the human being. The school places strong emphasis on values such as cooperation and selfless service to humanity, reflecting these in its educational practices and community interactions. The school’s philosophy posits that education must be viewed within the broader context of societal changes, interpreting the current era in human history as a phase of evolutionary transition from childhood to maturity. It suggests that, in this transitional phase, outdated methods and behaviours must be shed to develop virtues, powers and capacities suited to a mature society. According to Adam, the school supports the notion that humanity and cooperation across global communities can be enhanced through AI by introducing students to diverse perspectives and global issues through technology-enhanced curricula. To that end, Adam felt supportive of investigating AI tools in his classrooms.

It is also important to acknowledge that the context of this study may embody inherent privileges and structural features that do not fully align with those of public or universally accessible K-12 schools. However, due to its unique context, Adam and I both believe that his school provided a valuable opportunity to critically examine the practical application of democratic ideals (within its possibilities) using GenAI, which may be challenging or unfeasible in other school contexts. Democratic learning is fundamentally about pedagogy, which prioritises participation, inclusion and enhancement within the classroom. These values can be implemented regardless of the exclusivity of the school setting. While access to the institution itself may be limited due to socio-economic barriers, the pedagogical practices within the classroom can still embody certain democratic principles by fostering student agency, collaborative decision-making and critical thinking.

Findings

The data analysis revealed several themes relating to the research questions. For the first research question, three themes emerged: (a) AI-assisted lesson planning, (b) shifting learning objectives, (c) enhancing creative processes and (d) closing the gaps with personalised education. For the second research question, an additional three themes emerged: (e) limited space for integration of GenAI tools in the music curriculum, (f) the evolving and deterministic nature of GenAI tools and (g) lack of critical agency from the learners.

RQ1: How do GenAI tools shape and influence the pedagogical strategies employed by the music educator?

AI-assisted lesson planning

Adam reflected that GenAI tools streamlined his lesson planning process, allowing him to quickly generate and adapt lesson plans based on his needs, which enhanced his ability to focus on interactive and engaging classroom activities. Adam noted,

Adam also stated:

Adam appreciated the efficiency GenAI brings to lesson planning. The ability of these tools to quickly generate and adapt plans according to specific educational needs allowed teachers like him to redirect their focus towards more dynamic and student-centred activities in the classroom.

Shifting learning objectives

Adam mentioned how GenAI tools can be powerful for novice students ‘to create something that is super high quality’ with ‘very little to nothing’ in terms of prior knowledge in music. He claimed,

Adam emphasised that learning objectives can thus be recalibrated to foster confidence and creativity and encourage experimentation among novices. During my observation of Adam’s teaching, it became evident that he has actively shifted his learning objectives to align with the capabilities of GenAI tools. Rather than focusing on technical mastery of music production software, Adam emphasised creativity, exploration and practical application. For instance, he guided students in using Soundtrap, an AI-powered digital audio workstation (DAW), to compose original pieces, encouraging them to experiment with different genres and sounds without fear of making mistakes. He framed the tools as enablers of artistic expression, allowing students to achieve results that might have been unattainable through traditional methods.

Enhancing creative processes

Adam showcased one incident on how GenAI can be tailored to support specific educational outcomes while attempting to maintain students’ oversight and creativity within the AI-assisted learning process. He illustrated,

Perhaps due to Adam’s emphasis on original work, the students did not adopt the AI-generated melody directly. Instead, they reportedly used it as a starting point or inspiration for their own creative output. This approach highlights that, in this case, the AI reportedly served as a tool to enhance student creativity without replacing it.

Closing the gaps with personalised education

Adam recognised the significant potential of GenAI for personalised education when utilised well. He emphasised:

Adam also commented:

However, Adam noted that implementing GenAI effectively in music classrooms presented challenges. Most AI tools used were not specifically designed for music education contexts, which complicates their application. While these tools could assist students in preparing for exams such as the music IGCSE examinations or project presentations, GenAI’s impact was less significant in the music classroom. Furthermore, Adam acknowledged the difficulties in achieving democratic education, whether AI-assisted or not. These challenges are further elaborated on in the discussion of RQ2.

RQ2: What are the main concerns and challenges when integrating GenAI tools into democratic music teaching practices?

Adam actively fostered democratic learning within his rock band classes, emphasising student agency in song selection, small group collaborations and cooperative rehearsal strategies. Nevertheless, the integration of GenAI tools among students remained constrained at this stage, limiting its capacity to promote a truly democratic learning environment. These limitations were particularly evident in areas involving the limited space of integration of GenAI in the music curriculum, the evolving and deterministic nature of GenAI and the lack of critical agency from the learners.

Limited space for integration of GenAI tools in the music curriculum

Although there were few demonstrations of democratic values observed using GenAI tools as exemplified by Adam, he admitted the use of GenAI was not regularly integrated into the classroom curriculum. Instead, students were encouraged to engage with these tools during their afterschool hours. While Soundtrap is sometimes embedded in the classroom, it is currently more of an AI-assisted tool, with most of its functions not being AI-driven. This also had to do with the current lack of accessible AI tools for his K-12 music classroom.

In addition, while he was a supporter of democratic music learning, he believed the external pressure from administrators for testing made it difficult for him for implementation. He noted,

The concept of democratic education is centred on empowering students to take control of their learning processes. However, as students mature, nurturing this autonomy becomes increasingly complex. While younger students may engage more readily in exploratory and self-directed learning methods, the rigid structure of traditional education systems often suppresses these tendencies as they grow older. Furthermore, Adam stressed that the emphasis on exam preparation frequently fostered a ‘teaching to the test’ culture that prioritised test results over true exploration and intellectual curiosity. The integration of GenAI tools further complicates the pursuit of democratic education, making it even harder to implement effectively. As a result, democratic education remains an aspirational ideal for many teachers like Adam, yet it is challenging to achieve fully in practice.

The evolving and deterministic nature of GenAI tools

At this stage, though GenAI tools evolved at a rapid speed, Adam believed that the power of GenAI tools is still limited and cannot be compared to a real human educator.

Adam also reflected that while GenAI tools hold potential, many are not widely accessible and are not specifically designed to address the needs of music learning. For instance, from his experience, most GenAI tools he encountered were overly ‘generative’ in nature. Adam observed a significant limitation in the current generation of AI tools used in music education, particularly in their deterministic nature. Deterministic AI refers to systems that operate based on predefined algorithms and datasets, producing outputs that follow specific patterns without allowing for meaningful variability, spontaneity or deep user agency. In the context of GenAI tools, this determinism manifests in how these tools generate music or assist with composition – often relying on fixed patterns, pre-trained models and automated decision-making processes that lack adaptability to individual learning needs and creative exploration.

Lack of critical agency from the learners

Adam stated that the biggest issue about agency in using GenAI tools comes from permission from teachers and school policy. It was fortunate that his schools were not against the usage of AI and focused on having students utilise it with project-based learning. However, Adam reflected that students sometimes do not employ critical thinking while using GenAI tools; instead, they only copy and paste directly.

This highlighted a prevalent issue in digital literacy, where students often treat AI like a basic search engine, quickly accepting the first answers they encounter. This tendency among digital natives to prioritise speed over depth in their interactions with information technology reflects a broader cultural shift. In addition, he worried that the overuse of AI-generated music could hinder the development of a student’s individual creative voice. Students might rely too heavily on GenAI outputs rather than exploring their unique ideas, which conflicts with the intended purpose of using GenAI tools to inspire creativity. Nevertheless, Adam strongly voiced that while there is potential for misuse, this fear of misuse should not bar students from using AI tools.

Implications, limitations and conclusions

Under the democratic principles, GenAI can be conceptualised as a conduit to ‘sacred’ knowledge, for example, insights and understandings that might otherwise remain obscure or even inaccessible (Kucirkova & Leaton Gray, Reference KUCIRKOVA and LEATON GRAY2023). However, authentic democratic learning necessitates a degree of participation, not merely at a surface level, but at a deeper cognitive level, which remains dubious within the current frameworks and deterministic nature of GenAI. Authentic democratic learning in music requires users to engage deeply with their creative processes, question assumptions and explore multiple interpretations and solutions. Therefore, music educators can utilise AI as a starting point but not as an end product. For example, students could analyse and compare AI-generated pieces, discussing the artistic choices made and the implications for different genres. Another example could be using AI-generated compositions as a starting point for students’ own creative projects. Students could take an AI-generated melody and develop harmonies, rhythms and lyrics around it. Nevertheless, GenAI holds appreciable potential in assisting educators by automating certain instructional tasks, such as generating information, music histories, and rubrics, thereby potentially allowing more time to be allocated towards profound, student-centred learning. Furthermore, music educators should consider adapting their curricula to include AI literacy, guiding students to use AI as a creative collaborator rather than merely a shortcut to answers. As GenAI is here to stay, a more intentional curricular approach is necessary to ensure that students not only use AI tools effectively but also critically examine their limitations and ethical implications (Lee, 2025).

I must also recognise the limitation of this case study. This intrinsic case study is presented explicitly in Denzin’s account of ‘interpretive interactionism’. In this approach, ‘Each person, and each relationship, studied is assumed to be a universal singular, or a single instance of the universal themes that structure the postmodern period’ (Denzin, Reference DENZIN1989, 139). Yin (Reference YIN2009) argues that what a case study amounts to is a commitment to studying cases in their own terms, rather than in terms of prior categorisations: to documenting their uniqueness. Though I cannot generalise this experience to all other cases, Adam, as a universal singular narrative, holds his power to testify to his own words. With the rapid changes of AI, this also means this study serves this particular truth at this particular time of conducting and writing the study. Perhaps by the time this study gets published, AI tools might have evolved. Further studies on this topic will need to be conducted. Moreover, the accessibility and availability of GenAI, particularly with respect to music tools, remain significant issues that are underdiscussed in this study.

This study advocates for a holistic and balanced approach to AI integration in music education, promoting both technological advancements and the preservation of democratic pedagogical principles. In the near future, perhaps GenAI can be developed in ways that genuinely empower students to actively shape their learning experiences, provide meaningful feedback and maintain agency over their creative processes. Rather than outright rejecting AI or adopting it unquestioningly, it is incumbent upon music educators to continue exploring best practices within their domain and thoughtfully integrate AI tools into their classrooms. The AI education era is inevitable, and we must prepare to meet its challenges and opportunities with careful consideration.