Using ChatGPT for “children’s homework” is a good idea or not… we asked a data science expert.

Should we forbid it because it’s cheating? Should we teach them to “think about what they can do with it”

In addition to the spring break homework problem, many parents may want to have their children review the end of the school year before entering or moving on to the next grade.

The question we would like to consider is whether or not it is a good idea to use the popular “ChatGPT” to do homework.

ChatGPT is a chatbot developed by OpenAI in November 2010. The company has already been actively discussing whether ChatGPT can be used for university students’ theses and business people’s reports, etc.

So what are the advantages and disadvantages of using ChatGPT to do children’s homework?　As a member of the World Economic Forum (WEF) and one of the advocates of Sustainability Studies for applying the SDGs to education and management, Kento Sasano, a sustainability scholar who leads a genius group researching and developing new SDG/ESG data science technologies, says.

‘The benefits of ChatGPT are unmatched. Convenience. For example, it can be used at the level of the second examination of the University of Tokyo.”

In fact, I tried a world history question that asked, ‘Explain in two lines or less the claims made by Machiavelli in his “The Theory of Monarchs,” and I got a high enough score at the high school level.

Basically, I am good at writing and summarizing, and I can do creative writing such as writing a diary and writing a novel. However, there is the problem that they have no sense of ethics, so they don’t know what they should not do.

As a test, when I asked him to “write a diary of a day during summer vacation in the third grade,” the following sentences came out.

“ Today is one of the most fun days of my summer vacation! In the morning, my family and I decided to go to the beach.”

We woke up early in the morning and got ready. My mom prepared onigiri (rice balls) and sandwiches, so we took them and drove to the beach. The drive was fun.

Once we arrived at the beach, we quickly changed into our swimsuits and went out on the beach. The water was cool and refreshing. My father brought a floatation ring, which made playing in the ocean even more fun.

For lunch, we ate onigiri (rice balls) and sandwiches that we brought with us. My mom also prepared cold juice for us, so it was very delicious to eat near the ocean. We spent the afternoon playing on the beach. We played in the ocean, in the sand, and played catch with my dad. In the evening, we ate delicious shaved ice and heard that there would be a fireworks display at night.

“We are having a great day, playing a lot during the summer vacation and having a lot of fun with our friends. I am looking forward to tomorrow as well! “

The author has written a detailed description of his family’s leisure activities based on the proper elementary school and summer vacation. The composition is also well written. Furthermore can be written in Kanji characters, but playing a lot is duplicated, and “I’m looking forward to tomorrow too! The book is well structured, and although there is a slight imbalance in the way the author can write more in Chinese characters but duplicates a lot of fun and concludes innocently with “I’m looking forward to tomorrow, too!”

Danger of not being able to think and not being able to acquire literacy

What are the do’s and don’ts of using ChatGPT for homework?

First of all, at this point in time, you cannot input longer than about 2,500 characters. First of all, at the moment, it is not possible to input longer sentences than about 2,500 characters.

On the other hand, since ChatGPT cannot be used for papers that are that large in terms of structure, it is safe to say that it will not lead to serious misconduct.

However, since the system is designed in such a way that engineers can freely use it if they are paid through API collaboration (using APIs to link data functions between applications or systems to expand the functions available), it is probably only a matter of time before it is improved and longer sentences can be created in the future.

On the other hand, this is also a danger, and it is fair to point out the dangers of not being able to think clearly or acquire literacy skills when using the system for homework.

ChatGPT uses OpenAI’s GPT-3.5 model, which is trained by a 40-member “AI design team” of contract workers using reinforcement learning (RLHF) in the form of writing sentences such as, “This is how I would answer about 00,” and responds in a human-like manner.

The Japanese version is a Japanese translation of the original made in English, so the translation takes up to 20 seconds longer than when the questions are asked in English.

ChatGPT is a system that takes answers from all kinds of web pages and composes them for you, but homework should be about learning by reading and writing while worrying, and then correcting and learning as you have it corrected.

Besides, many parents talk about the importance of making mistakes, but if you use ChatGPT to give a wrong answer, you can use the excuse that ‘it was just derived by the AI.

You can’t tell them, ‘You can’t lie like that,’ or ‘You can’t take someone’s text without their permission.

The lesson will be to prompt engineers to provide highly accurate answers.

On the other hand, Mr. Sasano talks about the differences from conventional search engines and their possibilities.

In the case of Google search, the unit of search was a word, but with ChatGPT, it is now possible to search by sentence or context. This is not just a change in the business model, but a big difference in the way searches are conducted.

I think this is “the possibility for humanities majors to open the way for their activities in the engineering field.”

That is why we should utilize it with an understanding of its dangers, and if we don’t get on board, I think it is one keyword that will swallow us up when we enter society.

Regarding usage, Mr. Sasano focuses on prompts (incantations), i.e., the sentences used to ask questions in ChatGPT.

The sentences that come out are completely different depending on how you ask the question. The prompts are what make it possible to get the answers you are looking for on ChatGPT with a high degree of accuracy, and what is being tested right now is language skills.

I think there will be a demand for prompt engineers, an area in the engineering community where humans help AI think.

“Even with cutting-edge technology, whether the output will be interesting or boring and conventional will ultimately depend on the writing ability of the person using it, so it is necessary to combine the listening skills well to come up with good answers, and if that gets the homework done, then that’s good. Yes.”

Two rules necessary for homework use

In using them, it is of course necessary to put rules on them.

One is to make the source or sources clear.

It is natural that a researcher’s paper will always be a text taken from somewhere on the web, so it would be good if each family could be instructed to be very verbal about always asking whose paper it is and where the article is sourced from, especially when explaining facts.

Another thing to remember is that ChatGPT can ‘get it wrong’.

For example, when university students use ChatGPT for reports, they should use it to search for 10 articles to read for each theme or term as a rough preliminary research.

Since ChatGPT can also create tables, it is possible, for example, to create quizzes for each unit of five subjects in preparation for a junior high school mid-term exam.

However, the educational world still does not know how ChatGPT will affect children’s development, or at what age it is safe to start using it.

Mr. Sasano hopes that the Ministry of Education, Culture, Sports, Science, and Technology (MEXT) and boards of education will take the lead in the project to create guidelines in the future.

Digital slaves or digital rule makers

I have no doubt that the existence of ChatGPT will change entrance examinations in the future.

It will not end up being a debate about what is and is not allowed to be used, and in the future, it will not be a question that can be easily answered by ChatGPT, but more In the future, there will be more questions that make you think, questions that make you read the material and say what you think, and so on.

By the time the generation of digital natives and AI natives grow up to take college entrance exams, it will be almost certain that the basic premise of entrance exams will be to ask questions for which ChatGPT is rather commonplace to use, and for which the answers are still not available.

When children use ChatGPT for homework, there are some points that parents should be aware of.

AI can only isolate or collect data and turn it into information, but it cannot turn information into knowledge, which is the domain of human beings. This is a problem area for us humans.

I often use the analogy of “thinking you have a friend inside who knows things,” and when you ask them what you want to know, they do their best to make it up, but they can also make mistakes. But through communication, the correct knowledge increases.

Instead of simply stopping them from ‘not using it,’ I think it’s good for parents to try it out together, play with them, and make guidelines and family rules for using it.

Now is the transitional period between being a digital slave and a digital rule maker. Parents and children can communicate together and research their interests. Or we could try planning a family vacation together.

Kento S asano, sustainability scholar, SDGs communicator, and businessman, is one of the proponents of sustainability studies to apply the SDGs (Sustainable Development Goals) to management and education. He wandered the borderline between life and death during his high school years and was awakened to philosophy. After participating in the International Philosophy Olympiad and studying at the Kyoto University Faculty of Law and the University of Tokyo Graduate School of Interdisciplinary Information Studies, he became the Director of the Institute for Sustainability Research (Sustainability Research Institute) and Representative Director of Scrumy Inc. He is currently engaged in research and studies for the realization of a sustainable society and the development of technologies for SDGs management and SDGs education programs.