Utilize generative AI in moral education classes

0. Introduction

It’s been ten months since I wrote my previous article about the educational use of AI. During this time, ChatGPT, which I use in my AI research, has seen a shift from the GPT-4 model to the now mainstream GPT-4o model. The response time has significantly decreased, and the usage fees through the API have been greatly reduced, making it more accessible for editing long texts and provide feedback on the students’ answers in bulk.

In this article, I will introduce a case of using GPT-4o in middle school moral education as an example of utilizing generative AI in education. In the subject of morals, differences in how each student perceives the topic and their values become prominent.Therefore, it has become a difficult subject to teach in Japan.This article explores how much the latest AI can be utilized in the field of moral education.

In this article, the analysis of the questions asked and the corrections made to the students’ work were done using GPT-ScoreAID, which is provided by this site. GPT-ScoreAID has a feature that allows for the bulk processing of corrections and grading of texts written by multiple students, and we are utilizing this feature.

1. The Paradigm of Moral Education Classes in Japan (For international readers. Japanese readers may skip to the next section.)

Currently, moral education classes in Japanese middle schools, textbooks are often used even in moral education classes. In some cases, teachers create custom materials to address specific issues faced by the students or the local community, providing special topics for discussion.

Schools that center their moral education classes around textbooks often use supplementary materials such as a “Morality Notebook” edited by the textbook publishers. This notebook includes spaces prepared for students to fill out during class in accordance with the lesson, where they can write down their thoughts, insights gained from discussions, and other things they have noticed.

The National Course of Study in Japan, issued by the Ministry of Education, Culture, Sports, Science and Technology, outlines 22 content items for moral education at the middle school level. Schools create annual teaching plans consisting of 35 hours to cover these 22 items, taking into consideration the students’ needs and local issues.

A typical one-hour moral education class begins with reading the main topic or material in the textbook, which serves as the basis for discussion. Students then share their individual opinions and thoughts on the subject. This discussion is aimed at helping students appreciate the diversity of perspectives and values that may differ from their own, fostering learning through peer exchange rather than imposing values from the teacher.

In each moral education material, at the most important part of the lesson, there are “central questions” prepared. Through these questions, students will contemplate the moral values embedded in the material.These questions are often provided as examples by the textbook publishers in the teaching resources. Many teachers use these resources to guide their classes, introducing the “central question” at crucial points to prompt students to reflect on and discuss the core message of the material. Teaching resources may also include examples of supplementary questions or instructional methods to help students understand the protagonist’s feelings or capture the context of the material.

However, given the emphasis on discussions among students and valuing diverse perspectives, these prepared questions may not always function effectively. Teachers may need to adapt and adjust their approach in real-time to ensure productive class discussions. Therefore, the current state of moral education in Japan requires a certain level of competence from the teachers in charge in order to effectively conduct moral education classes.

2. About the Moral Education Material Used in This Report

In the class, we used the moral education material for first-year middle school students titled “I Will Become a Professional Cleaner” (published by Nihon Bunkyo Publishing Co.Ltd.). This material follows the life and changing attitudes towards cleaning work of Haruko Niitsu, an environmental master of Haneda Airport, which is renowned as the cleanest airport in the world. The class revolves around a central question: “Your cleaning lacks heart,” a statement from a cherished mentor who nurtured Ms. Niitsu. This serves as a launching point for exploring topics such as the essence of work, the joy and purpose of working, and one’s connection to society.

Traditionally, moral education classes involved students sharing their opinions and thoughts through raising hands, with these ideas being collated and summarized on the blackboard. However, recent classes have increasingly incorporated ICT, utilizing applications like Figma and Canva to compile student thoughts on sticky notes, making these visible for discussion.

In the current approach, to integrate AI into the learning process, students were asked to respond to questions via Google Forms during the class. The responses were collected in a spreadsheet and then analyzed and corrected by ChatGPT through GPT-ScoreAID. This process helped identify perspectives that could stimulate lively discussions and also provided advice tailored to individual students.

With the capabilities of current AI, real-time APIs have become more accessible. While we used forms as input in this instance, it may soon be possible to have AI directly listen to student discussions and provide guidance on the discussion’s direction. If AI can be seamlessly integrated into the natural flow of a lesson, its use in educational settings might become more commonplace.

3. Classify students’ responses using AI to find better directions for discussions in class.

After reading the main text, I conducted the first inquiry as outlined in the teaching plan suggested by the textbook supplier.

Question 1: “Understanding Emotions ①” — What do you think were Haruko Niitsu’s feelings when she came to Japan from China and started working in cleaning?

The student responses recorded in Google Forms were as follows:

• “Why such a job?”
• “I can’t imagine what could possibly be enjoyable about doing cleaning work.”
• “Feeling a bit concerned as it was different from what I expected.”
• “Wondering what kind of work cleaning is. Feeling nervous.”
• “I need to endure until I get used to Japanese because when I was in China, I was well-off.”
• “I think I was confused when I suddenly came to Japan and started cleaning work.”
• “Cleaning is something you do when you can’t speak the language, not because you want to do it.”
• “I didn’t really want to do it, but I had no choice because I need to earn money.”
• 和訳して下さい
• “I’m not a fan of mundane and unglamorous jobs like cleaning, but I need the money, so I’ll put up with it for now.”
• (and so forth)

I compiled these responses and inquired ChatGPT (using the GPT-4o model) to categorize the students’ answers and identify 2–3 perspectives likely to spark discussion. About 30 seconds later, ChatGPT responded with the following analysis:

ChatGPT: Upon reviewing the students’ responses, several common themes and perspectives emerge. These can be categorized as follows, and are likely to provoke discussion:

Concerns or Negative Impressions about Cleaning Work:

• “Why such a job?”
• “I don’t want to do it and dislike it.”
• “It might be a bit troublesome.”

In this perspective, many students have expressed concerns or negative impressions about cleaning work, providing an opportunity for discussions around the reasons and background for these opinions. Discussions can also explore what might change these impressions.

Concerns About Language and Environment:

• “Worried about not understanding Japanese.”
• “What should I do if a customer talks to me while I’m doing cleaning work?”
• “I don’t want to have to work instead of going to school.”

There are opinions concerning anxieties about language and environmental changes upon coming to Japan. This perspective provides an opportunity to discuss the challenges and anxieties of adapting to new environments, perhaps prompting students to consider how they would handle being in a similar situation.

Acceptance Despite Reluctance:

• “This is the only option, so I’ll do my best.”
• “I’m doing cleaning work because I can’t speak Japanese and it’s hard to find other jobs.”
• “No choice but to clean to earn money.”

Some students exhibit an acceptance of their situation and are taking a more proactive stance. This perspective can be used to discuss the significance of making efforts positively in challenging situations and the growth that can result from it.

The purpose of this initial question was to infer Niitsu’s feelings and her initial sentiments towards cleaning work. Based on the responses, we encouraged students to express their views from two directions: “concerns about the cleaning job itself” and “concerns about language and the environment they were placed in,” delving deeper into Niitsu’s feelings when she first took up cleaning work. After confirming Niitsu’s initial feelings collectively, the lesson moved to the second inquiry as indicated in the teaching example.

“Understanding Emotions ②” — After progressing to high school and meanwhile working at a sound company, Niitsu continued doing cleaning work during her spare time. How did her feelings towards cleaning change?

Student responses were as follows:

• “Maybe cleaning really suited me.”
• “I thought that cleaning was more enjoyable than working at the sound company.”
• “I used to do it normally, but now I’m interested in cleaning and my feelings changed.”
• “Initially unsure and unwilling, but found cleaning to be a very fulfilling job after trying it.”
• “I’ve always liked being physically active, so I thought cleaning suited me.”
• “I’m starting to see cleaning work in a new light. I wonder if I can really do it well for everyone’s sake? I’m worried.”
• “I found that cleaning work suited me as it helped to ease my tension and loosen up my muscles.”
• (and so forth)

ChatGPT identified the following three perspectives likely to spark discussion based on the responses:

ChatGPT’s Analysis:

The Fulfillment and Enjoyment of Cleaning Work:
Many students responded that “cleaning work is fulfilling,” “it’s enjoyable to make things clean,” and “there’s a sense of accomplishment.” This perspective allows for discussing how Niitsu found fulfillment and enjoyment through cleaning work, and exploring the elements underlying these feelings. Questions like “Why did she find cleaning enjoyable?” and “What provided her with a sense of accomplishment?” can be posed.

The Importance of Physical Activity:
Responses such as “I like being physically active,” and “It energizes me,” appeared frequently. This perspective allows for a deeper dive into the relationship between physical activity and emotional satisfaction. Discussions could explore, “Why was physical activity comforting?” and “Why did cleaning suit her physically compared to other jobs?”

Self-Discovery and Aptitude:
There were responses such as “Cleaning suited me,” and “I found cleaning better than the sound company job.” Here, we can explore how Niitsu realized cleaning suited her and went through a process of self-discovery. Students can be prompted to exchange ideas on topics like, “What experiences influenced Niitsu’s self-perception?” and “How did Niitsu transition to a cleaning career?”

In the second inquiry, three directions for discussion were proposed. Leveraging this, the class discussion focused on affirming how Niitsu found fulfillment and enjoyment in cleaning. Eventually, I steered the discussion in a direction where students could feel the flow of ‘self-discovery in finding a job that suits them.

4. Analysis of central questions by AI

Now we turn to the most crucial part of the moral education lesson: the central question. It wouldn’t be an exaggeration to say that the two preceding questions serve to pave the way for this central question. Let’s see if, through the two questions and discussions in this lesson, we were able to get the students to thoroughly consider the following central question.

“Central Question” — Let’s imagine how Niitsu felt when her mentor, Mr. Suzuki, told her, “Your work lacks heart.”

The responses from the students included:

• “What does that negative comment mean, when I’ve been working so hard in this cleaning job?”
• “I am indeed working with heart!”
• “Sad after all the effort I put in.”
• “Why, even though I clean places that aren’t visible?”
• “I’ve done so much cleaning work but what is lacking? What does ‘heart’ mean?”
• “What does it mean to not have heart?”
• “Despite working hard to keep things clean, why am I being told this?”
• “I work hard, so it’s a bit sad, and I don’t know what to fix.”
• “Why am I told this despite studying hard in vocational school, passing exams, getting employed, and giving my all in cleaning?”
• “I have the skills, but I don’t know what I’m lacking.”
• “I’m working hard, so what am I missing?”
• “Huh? I worked really hard though!”
• “I wondered why I was told ‘your work lacks heart.’”
• (and so on)

To facilitate an active discussion moving towards the instructional objective of “Diligence and Public Spirit,” based on the most crucial central question, we obtained advice from ChatGPT to classify student opinions and promote discussion from the following perspectives:”

Balancing Skills and Empathy:

Perspective: Many students are puzzled about being told they have skills but lack heart. Here, you can explore the importance of balancing technical efforts and mental attitudes.

Example: You can discuss themes such as “If skills alone aren’t enough, in what ways do we need to be conscious of others?” and “What exactly does it mean to work with heart?”

Self-awareness and Impact on Others:

Perspective: Among the students, there are many questions about “what’s lacking” and what heart means, considering their self-awareness and how they are perceived by others. Here, you can provide an opportunity to think about how one’s actions are seen by others.

Example: By raising questions such as “Have you considered how users or colleagues feel when you clean?” and “How do you want to be perceived by others?” you can facilitate discussions on the importance of being aware of others’ perspectives.

Personal Growth and the Meaning of Work:

Perspective: Questions like “I don’t know what’s missing in cleaning work” and “what’s lacking?” contain inquiries about personal growth and the meaning of work. Through this, you can encourage students to consider how they discover personal growth and the meaning of their work.

Example: Providing opportunities to discuss “What approaches do you think are needed for personal growth?” and “What joys can be gained through work?” can encourage students to think concretely about how work contributes to personal growth.

ChatGPT provided three discussion concepts. It also highlighted that the students could be divided into three layers: students who simply “don’t know what’s missing,” those who are exploring what might be lacking, and those who, in addition to their cleaning skills, are aware of the mental aspects of work or its societal value. The analysis of the responses reveals the presence of these three layers in response to the central question in this class.

Therefore, being conscious of these three layers present in the classroom, we intentionally created opportunities for students to hear opinions from peers in different layers or engage in discussions about questions and doubts they have. Through this approach, students were encouraged to broaden and deepen their understanding of what work means—examining its joys, purpose, and connection to society.

5. Utilizing AI to advise on individual student’s growth and changes in values.

In the previous chapter, I utilized it to collect and analyze students’ responses to determine the flow of the class. Here, I will consider another way of using AI, which is to track each student’s responses to questions, and use AI to observe what the student has learned from the material and how it has contributed to their own growth.

The current curriculum guidelines emphasize the importance of observing students’ growth and changes in values based on specific descriptions and statements in moral notebooks and presentations in class. Therefore, based on the responses to the three questions asked in class, I decided to have ChatGPT create advice for each individual student. I will post some of them. (The prompts will be detailed in Chapter 6.)

In this practice, I had AI generate draft advice based on the three descriptions each student answered in the moral education class, and I, as a teacher, checked and revised it. Then, I aggregated the confirmed advice on a spreadsheet, linked it with email addresses, and used a code called GAS to return it to each student by email.
In the next day’s class, when I told the students, “I sent you an email with my thoughts and advice on what you wrote in yesterday’s moral education class. Please read it,” the students read the feedback and advice returned by email with interest and seriousness.

In the case of using a paper “moral notebook”, the student’s thoughts written there will not spread to others unless they are presented in class. In classes that utilize ICT such as Figma or Canva, all students’ thoughts can be collected by sticky notes, allowing us to shed light on opinions that were previously overlooked. With this form method, we can capture what each individual felt and perceived, and with the help of AI, we can classify and stratify students’ opinions to utilize in the flow of the class. This has proven to be very effective in conducting moral education classes.

Also, in the case of individual corrections for students, it used to take about two hours per class just to check the descriptions on the paper moral notebooks and write a few lines of comments from the teacher. If there are after-school activities such as club activities, the inspection of the moral notebooks is postponed, which can be a real problem in workplaces that are promoting work style reform.
By utilizing AI in such situations to create draft advice for students, there is a possibility that the time required for corrections can be significantly reduced by the teacher using that draft to add and revise, and then return it by email.

In this practice as well, I was able to return the draft corrections created by AI with additions and revisions in about an hour. Above all, the correction process proceeds by checking, revising, and adding to the advice text that AI has analyzed, rather than drafting correction sentences from scratch. This has shown to be much more efficient than writing and returning on paper.

Until now, there have been many school sites where, amidst the struggle with time, they could only go as far as stamping “checked” on the moral notebooks. However, by introducing electronic moral notebooks, teachers can change their approach to work with AI to correct and advise on each child’s understanding and feelings. Despite the extremely busy educational field, AI is beginning to show its contours as a reliable assistant in schools, supporting and assisting teachers’ work.

6. About the Prompt

Finally, I would like to discuss the prompts, which served as instructions for the AI utilized in this practice. In generative AI, the quality of the prompt greatly influences the quality of the response. Below, I will provide the prompts used to create personalized advice for each student during this moral education exercise.

After reading the prompt, you might feel that it is quite detailed and comprehensive. When I devise a prompt, I first consider how I would instruct a human college student if I were to request similar revisions. In such a case, I would surely communicate details like: 1) what story forms the basis, and 2) what intentions the teacher has for each inquiry posed to the students.

My current view of AI prompts is that, similar to humans, AI cannot perform accurate revisions unless it is provided with appropriate information. This perspective is clearly reflected in the prompt shared here. Applications like GPT-ScoreAID, which leverage AI via API, can handle prompts of this complexity with ease. As the use of AI becomes more sophisticated, I anticipate that prompts will become more detailed and longer. Schools adopting this technology will need to grow accustomed to creating and using such detailed prompts.

Furthermore, when requesting AI to handle complex content, like in this example, it’s rare for a prompt to function ideally on the first try. A repeated process of slightly modifying the prompts and refining them until they produce the desired outcome by the teacher is often necessary. The GPT-ScoreAID used in this practice includes a function for refining such prompts, which I felt was essential for applications used in live lessons.

7. Conclusion

Moral education classes are one of the most demanding for teachers, requiring the ability to appropriately change the flow of the class based on the interpretation of teaching materials and interaction with students. By introducing generative AI, it is possible to carry out the practices introduced in this paper. Things that were not possible with previous generative AI are now becoming possible, and I am once again struck by the incredible speed of AI’s evolution.

In this practice, we used a model called GPT-4o from ChatGPT.
ChatGPT already has a successor model called o1 that can perform more advanced inference, and there is also mention of a further successor model called o3 aimed at AGI in January 2025. By 2025, I predict that there will be reports of practices using the o1 model and experimental initiatives using the o3 model in the field of education, and research on the utilization of AI will progress far beyond the level of this report.

In my report this time, I was unable to cover the utilization of the o1 model, but I intend to add features as soon as possible that allow the use of the o1 model in GPT-ScoreAID, and incorporate the functionality of the RealTimeAPI that can handle natural language.
In particular, the implementation of the RealTimeAPI feature will enable uses such as allowing the AI to directly listen in on student discussions and receive advice from the AI in a more natural flow, moving away from the current text-based input using forms. The threshold for AI is still high in the educational field, but an environment where interaction with AI is possible through voice input could be a major catalyst for promoting the use of AI in education.

As we welcome the new year, it is my New Year’s wish for 2025 that the practice of using generative AI in the field of education will increase even more, and that AI research in the field of education will further deepen and develop.

Toshiyuki Okuzaki, Former Principal of Hakodate Municipal Kita Junior High School

If you have any inquiries regarding this article, this is the place to ask!