A small number of essays generated by artificial intelligence were submitted to a School Library Association contest last year, but educators worry that without effective measures in place, such cheating could skyrocket.
The nationwide contest received more than 2.6 million essays on the theme of books read by elementary, junior high, and high school students during summer vacation.
Teachers noticed something was amiss with more than 10 essays, and the students who submitted them admitted to using the AI chatbot inappropriately, after which their essays were removed from the competition.
The next essay contest is scheduled for the fall, and officials are concerned that AI-generated work may further evade detection.
Students apply for the competition through their school and their essays are read by their teachers before they can move on to the next stage of selection.
The association will not completely ban the use of AI tools in its competitions.
“Our position is that generative AI can be educationally beneficial depending on how the technology is used,” a representative for the association said.
The association's guidelines for applicants prohibit plagiarism and inappropriate citations.
For example, an essay that contains content copied and pasted directly from AI-generated text would be an “inappropriate” use of the technology, the association said.
As AI chatbots have become so easily accessible to anyone, the association has been warning teachers since last year about the potential problems that can arise when using the tools.
The association has been conducting experiments to spot artificially generated text, but officials at the organization said AI technology is still ahead of detection techniques.
“Each time we sent the AI to write an essay, it produced something different,” the official said. “This made it difficult to distinguish the characteristics of the AI-generated material. Moreover, generation technology has improved dramatically since last year.”
The association said teachers, who are familiar with students' writing skills, are the most trustworthy people to check the authenticity of essays.
“If teachers can't find the cheating, I doubt anyone else will,” the official said.
Another essay contest aimed at high school students faces similar challenges.
Last year's contest received more than 20,000 essays, and organizers said in a post-contest survey that some judges said they suspected some of the submissions were created by AI.
The organizers said they could not say for sure whether the essays were created with AI because they had not been able to contact the essay's author, but the judges would not allow such essays to proceed to the next stage.
Organizers of last year's contest did not clarify their policy regarding computer-generated text, but they believe generative AI can be useful for brainstorming and proofreading.
At this year's competition, organizers plan to address the “inappropriate use” of AI models.
“Fairness is important,” said a source with the organization. “Some students highlight their winning essays during the high school and college admissions process in order to advance.”
But many organizers agree that current technology for distinguishing between human-written and AI-generated text remains unreliable.
In guidelines published last year, the Ministry of Education, Culture, Sports, Science and Technology said children from elementary school to high school should be aware of the importance of learning how to use AI tools.
However, he also noted that it was unacceptable for students to submit competition work that was “outsourced” to an AI.
Copied artwork
The use of AI art generators is also a concern.
Trident College, a vocational school in Nagoya that teaches production techniques for computer games, graphics, illustrations, etc., is accepting submissions for a 2023 2D and 3D illustration contest open to high school students across the country.
The use of AI models has been banned.
Our school is an affiliated school of Kawaijuku, a major cram school chain.
Trident University students with expertise in illustration were brought in to help weed out questionable submissions in the first round of judging.
Experts say that works created by AI generators tend to imitate existing works of art or specific artworks.
According to the school, none of the submitted works displayed characteristics that would indicate they were AI-generated artwork.
But school officials said human visual inspections may not be around for much longer.
“If users misuse AI tools and train them to create works that don't resemble specific works of art, they may not be able to recognize the content the AI generates,” said Yuji Kojima, a lecturer at the school.
Seeking solutions
Companies are working on programs to determine whether something was made by AI or a human.
Kawaijuku announced in March that it had developed a program to detect computer-generated text, part of a larger system it has been providing to universities since the 2018 academic year, primarily for the purpose of processing admissions procedures.
Kawaijuku said the new program could have broader applications, such as judging entries for essay contests and evaluating university students' reports.
When used to evaluate essay questions in admissions processes, the program is designed to issue a warning if it detects suspicious text.
“We feel there is a great need for tools that can recognize artificially generated text,” a Kawaijuku official said.
Citing university officials, the officials said AI may have been used to generate materials required for admission.
“Our program is only a guide, but may serve as part of an applicant's overall evaluation,” the official said.
Kazunori Sato, an associate professor at Shinshu University who specializes in media literacy and information education, emphasized the need for adults to use AI tools to understand such creative works.
“Essays produced by AI models tend to be bland and unoriginal,” he said. “Adults, including educators, should regularly challenge AI to learn patterns in computer-generated text.”
He noted that the use of generative AI is permitted in academia.
Sato said rules requiring researchers to report how and in what parts of their research they use AI are widely accepted to allow reviewers to carefully check for potential copyright issues.
The same approach can be applied to essay and other contests, he said.
Some contest organizers are even inviting submissions made with generative AI.
Last year, Gunma prefecture gave permission for works created by elementary, junior high and high school students using AI to be used in manga and anime events, computer graphics and game contests, etc.
While the prefecture has allowed the use of the AI model, it has warned students about the possibility of copyright infringement.
Entrants are obliged to declare which AI tools have been used in which parts of their work and in what way.
According to the prefecture, AI was used in 19 works, but none of them won awards.
A prefectural official said, “We believe that generative AI will become an indispensable tool in the future. We hope that by participating in events like this, students will develop their ability to use AI while minimizing the risk of copyright issues.”
(This article was written by Kohei Kano and Yukito Takahama.)
