During the age of generative artificial intelligence (GenAI), one concern among musicology instructors is the prevalence of AI-generated content in student essays. It is evident that GenAI, particularly ChatGPT and other Large Language Models (LLMs), has revolutionized the field of education. The development of GenAI policymaking in higher education is now closely intertwined with its practical implementation in teaching and learning. The increasing utilization of GenAI is an undeniable and irreversible trend that continues to shape technological progress in education. However, it also presents a significant opportunity for both teachers and students to reevaluate and enhance the learning process. While the use of GenAI may raise some concerns, it may also offer numerous educational advantages, such as its constant accessibility, potential alleviation of heavy workloads, and the ability to provide personalized one-on-one guidance. As Pierre Boulez once described, “Revolutions are celebrated when they are no longer dangerous.”
It is widely known that ChatGPT can serve as a co-designer of a course. It can assist in brainstorming ideas for course topics, developing tailor-made course outlines and teaching materials, formulating essay questions, and summarizing books. The generated materials can be further refined by posing follow-up questions using the Socratic method. On top of these, this blogpost suggests two additional ways in which ChatGPT can support the teaching of musicology courses. Drawing from my personal experiences in teaching Western music history and film music courses, I share how ChatGPT can support teaching and learning in musicology in two ways: by developing a tailor-made TA chatbot and by leveraging GenAI as a marking assistant. The suggestions emphasize the importance of human involvement and adherence to the guidelines outlined in UNESCO’s “Guidance for Generative AI in Education and Research” (2023) to adopt AI in a responsible way.
Leveraging a Custom Chatbot as a 24/7 Teaching Assistant
One of the valuable aspects of ChatGPT in education is its potential for personalized learning. The custom chatbot function in ChatGPT represents a powerful method for providing individualized support for students and facilitating self-directed study. By inputting course information, content, outline, deadlines, and readings into its “knowledge base,” the custom chatbot can serve as a self-directed revision and inquiry tool for a specific course. Users can also specify the chatbot’s role in the custom chatbot feature, designating it as a teaching assistant and requesting responses in a supportive and caring manner.
In my film music class, I used the custom chatbot feature of ChatGPT through poe.com to create a chatbot TA for my course. I transcribed my lecture videos into Word documents and uploaded them to the chatbot’s knowledge base, along with readings, the course outline, and assignment instructions. As a result, this TA chatbot possessed comprehensive knowledge about the course. I also instructed the chatbot to refer only to the materials I provided. If asked about something beyond the scope of the course, the chatbot should respond with “I don’t know.”
When presented with general questions about the course and its content, the chatbot demonstrates its ability to provide accurate answers, citing the relevant sections from the original sources as necessary. For example, if I ask, “Who is Joe Hisaishi?” the chatbot synthesizes information from all relevant portions related to Hisaishi, a Japanese film music composer, in the lecture transcripts and offers a coherent response. It also includes footnotes that users can refer to in order to locate the sources in the documents. As a result, the chatbot’s answering style remains consistent with the general function of ChatGPT, but is now limited to the course materials I provided.
Note: The author accessed ChatGPT via poe.com, a free platform for users. The website poe.com features the ChatGPT Large Language Mode, instead of the ChatGPT platform by OpenAI. Users in the US can build their custom chatbot directly in ChatGPT, which is available on the OpenAI platform. The chatbot then can be linked to a Learning Management System such as Canvas or Blackboard.
In fact, the inherent inference and reasoning capabilities of ChatGPT are often underestimated. These capabilities enable the chatbot to analyze and restructure data to answer questions. For instance, even though I did not explicitly outline the criteria for good film music in my lectures, when I asked the chatbot, “What makes good film music?” it analyzed various segments of the sources to deduce a plausible answer. It referred to different sections discussing the functions of film music and included bracketed footnotes that reference the location of the lecture transcripts:
Good film music is not just about creating beautiful melodies or catchy tunes. It serves a greater purpose in enhancing the overall cinematic experience for the audience [1]. One important aspect of film music is its ability to convey meaning and emotions in relation to the story and visual elements of the film [1]. It helps the audience engage with the movie and adds depth to the narrative [1] […]
Furthermore, effective film music should create a sense of continuity, seamlessly connecting different parts of the story [2].
Similarly, when I inquired about the characteristics of bad film music, the chatbot provided qualities that oppose the functions of film music. It mentioned aspects such as music that distracts from the narrative, poorly aligns with the scene’s mood, inappropriately incorporates music from different cultures or time periods, and lacks originality. It even searches relevant examples from the lecture transcripts. The custom chatbot seems to go beyond being a simple search engine for retrieving information, demonstrating itself to be an advanced engine capable of analyzing text to handle more complex inquiries.
With this course database, teachers can request quiz questions and teaching materials. More importantly, sharing this self-directed revision chatbot with students provides a platform for students who may feel hesitant to approach an instructor or teaching assistant with their questions—or even minor inquiries. Instead of scheduling appointments with the instructor or teaching assistant, students can consult the TA chatbot whenever they wish, without any time constraints. It is especially beneficial in larger classes where instructors may find it challenging to individually assist every student. This kind of virtual tutoring can also free up instructors’ time to improve other aspects of teaching and learning.
The platform does not offer an option to track user interactions, as students access the chatbot using their own accounts. However, students self-reported that after using this chatbot in the film class, they generally found it convenient, particularly for clarifying key terms and summarizing concepts drawn from various relevant sources. They also utilized it for course specifications and assignment information. At this stage, the teacher, as the creator of the chatbot, has no way to view students’ questions. The platform could become even more beneficial if this feature is implemented in the future.
Enhancing Assessment-Marking with GenAI Assistance
ChatGPT can also help alleviate teachers’ workloads by assisting with assessment marking. By functioning as a teaching assistant, ChatGPT can be utilized to conduct initial screenings and offer feedback suggestions to the teacher. ChatGPT has good reasoning capabilities when provided with clear instructions; when detailed grading rubrics are supplied, it can assist teachers in evaluating written assignments.
For instance, in the film music course I taught, I assigned a final essay proposal requiring students to submit a one-page document. The marking criteria for this assignment consisted of three components: 5 marks for the formulation of the research question, 5 marks for the demonstration of subject knowledge, and 5 marks for the quality of communication, totaling 15 marks. To facilitate the marking process, I incorporated the marking rubrics into the custom chatbot I developed earlier. The chatbot can refer to the course content while analyzing the assignments. Below is an example of how ChatGPT could provide marking feedback according to the rubrics:
ChatGPT is effective at marking straightforward assessment questions, such as multiple-choice and short-answer questions that require keyword identification. However, for essay-type assessments, the quality of marking heavily depends on the clarity and detail of the rubrics. It may not provide a fair evaluation of students’ innovative ideas or content that falls outside its knowledge base. Additionally, there can be variations in its analyses, leading to fluctuations in marking. For an essay scored out of 100, these variations can be significant. Therefore, it is advisable to use ChatGPT as an initial screening tool for suggestions or as one of the markers for the teacher’s consideration. Final grading should always be double-checked by a human.
While ChatGPT or other Large Language Models are not yet perfect, their rapid development holds great promise. By optimizing language processing capabilities and platform features, they have the potential to serve as valuable teaching assistant chatbots, enhancing the instruction of text-based music courses in the future.
