Universities are struggling over how to deal with generative artificial intelligence, such as ChatGPT, which could be used by students for cheating purposes.

AI has advanced to a level that makes it difficult for educators to differentiate between reports written by students and those generated by computer.

Many universities are restricting the use of generative AI in writing reports or are calling on students to be extremely careful about utilizing it.

However, the schools have recognized the benefits of generative AI for such purposes as collecting and sorting information, so many universities are avoiding outright bans.

The key is striking a balance.

ChatGPT, a service provided by U.S. startup OpenAI, can answer questions or respond to instructions in natural-sounding sentences.

Use of the chatbot has skyrocketed since the company made it available to the general public in autumn 2022.

It not only can carry on natural-sounding conversations, but it can also propose text structure, as well as translate, modify and summarize sentences.

And it is capable of writing computer code.

In March, OpenAI released GPT-4, the latest version of ChatGPT, for a monthly subscription fee.

Its improved abilities have sparked awe and concerns.

Kyoto University Provost Nagahiro Minato stressed the negative aspects of generative AI applications in his speech at the university’s entrance ceremony on April 7.

He pointed out that AI can generate sentences that contain incorrect information, and that AI is unable to correctly cite sources to support its arguments.

“AI can only search information (to collect it) and lacks the ability to examine it, which is research. That is a catastrophic shortcoming in the use of AI for writing academic reports.” Minato said in the speech.

At the end of March, Sophia University issued a notice for its teaching staff and students on using AI to write reports or theses.

It said the university will reject reports or theses that are generated by AI.

The notice also said the university will strictly apply its disciplinary rules about misconduct in response to confirmed AI-generated papers.

“However, with permission from teaching staff, students can use AI within the limits allowed by instructions,” the notice said.

It concluded, “We will continue reviewing and considering our approach to AI, including its appropriate use in education.”

The University of Tokyo on April 3 released a document of its views on the use of generative AI under the name of Kunihiro Ota, board director and vice president of the university.

The document referred to the possibility of AI-generated text including false information.

To make a good use of generative AI, “it is necessary to critically check and modify answers (generated by AI) as appropriate,” the document said.

“It’s our basic presumption that students write reports themselves, and they are not allowed to prepare reports only by using generative AI,” it said.

Ota told The Asahi Shimbun that when he himself used ChatGPT, he realized the chatbot could have a significant impact on students’ reports and their grades. He said university students have already started using it.

“GPT-4 can mark almost full scores on exams and report assignments that we have required for our students,” he said. “AI’s abilities have improved daily, and we can’t escape from the changes of the times.”

University of Tokyo officials started discussing how to approach generative AI in January.

They concluded that it would be difficult to prohibit students from using generative AI while they adjust to the changing times.

A consensus was reached within the university that it should use generative AI in a way that is beneficial for university education.

In line with this agreement, the university’s April 3 document said: “We believe we should act in anticipation of the various societal changes that generative AI will bring. We should actively try to find better ways to utilize generative AI, as well as new technologies and new legal, societal, and economic systems (enabled by generative AI.)”

Lui Yoshida, associate professor of educational technology at the University of Tokyo, studies the use of ChatGPT for educational purposes.

“(With ChatGPT’s emergence,) university teaching staff now need to be creative in designing assignments for students,” Yoshida said. “Also, the actual process of learning will be regarded as important, in addition to reports and exams. The focus will be on not just ‘what students have learned’ but also on ‘how they have learned.’”

Tohoku University in March published instructions for its teaching staff on how to deal with generative AI.

“It’s not realistic to completely ban the use of generative AI,” the instructions said.

But it mentioned problem situations, such as when “generative AI provides more correct answers than in the reports that students have seriously worked on.”

The document advised giving oral exams to students instead of assigning written reports.

It also said, “It is desirable that teachers themselves think about how to approach generative AI, such as whether to partially ban it or to provide instructions to students on how to use it.”

Showa Women’s University told a meeting attended by all teaching staff that if they want students to write reports without using AI, they should require the students to compose the reports at exam venues.

Some measures have already been introduced at Keio University.

Students at some seminars will be required to take oral exams at the end of the academic term when they submit reports.

In addition, some classes at Keio scrapped computer programming assignments because generative AI can write programming code.

One junior at the economics department at the university said he was asked to correct the answers generated by AI during a class.

“You can’t figure out if such answers are wrong unless you study sufficiently,” he said. “It was an innovative drill.”

(This article was written by Hajime Ueno, Takahiro Takenouchi and Senior Staff Writer Fumio Masutani.)