AI writing still not as good as real students: lacks personal experience and nuanced critical thinking, study says

May 28 news, according to foreign media SciTechDaily 26 reports, the University of East Anglia in the United Kingdom, the latest study pointed out that the level of artificial intelligence writing still has not caught up with the performance of real students. The results have been published in the journal "Written Communication".

AI writing still not as good as real students: lacks personal experience and nuanced critical thinking, study says

The researchers compared essays written by 145 college students to those generated by ChatGPT. While the AI-written essaysClearly structured and grammatically correct, but lacks a "human touch".-- a.k.a. personalized perspective. The lack ofPersonal experience and subtle critical thinking, precisely between it and student workclearest difference.

With generative AI capabilities rapidly increasing, the research team cautions that the education system should place greater emphasis on developingCritical Literacy and Ethical Awareness. They hope that this research will help teachers identify machine-generated content and thus tackle academic dishonesty more effectively on a global scale.

Ken Hyland, a professor at the university's School of Education and Lifelong Learning, said, "After ChatGPT was launched, a number of teachers began to worry that their students would use it for homework. The concern is that an AI tool like ChatGPTMay encourage cheatingthat undermine the literacy and critical skills that students are supposed to develop. Especially now, we lack the tools to reliably recognize AI writing. Therefore, we wanted to understand just how well AI can mimic student writing, and in particular, how well itThe ability to create interactions with readers like humans. "

In total, the researchers analyzed 145 college essays and 145 automatically generated articles by ChatGPTWe paid particular attention to so-called 'interaction markers', such as asking questions or adding personal comments," says Hyland. The results showed that the students' essays were richer in this area, interacted effectively with the reader and were more persuasive. They made good use ofRhetorical questioning, personal interjections, and direct reader statements -- All of these writing styles enhance expression, make reading connections, and build strong arguments."

"In contrast, ChatGPT's article, while fluent in language, wasIt seems detached and calm.It follows the norms of academic writing but does not show the author's personality or a clear position. While it follows the norms of academic writing, it fails to show the author's personality or take a clear stance.AI essays hardly ever ask questions and rarely incorporate personal opinions. As a whole, this type of article lacks impact, and the argumentation is much paler. This is precisely because it relies on statistical learning with a training corpus, whose core goal is to maintain linguistic coherence rather than pursuing changes in emotion and tone in dialog."

However, the study does not consider AI useless. Instead, the researchers suggest that tools such as ChatGPT be used to supplement instruction, rather than being a shortcut for speculation.

Hyland concludes, "Students enter school not just to learn to write, but to learn to think-that's whatNo algorithm can replaceof core competencies."

With links to research papers:DOI: 10.1177/07410883251328311

statement:The content of the source of public various media platforms, if the inclusion of the content violates your rights and interests, please contact the mailbox, this site will be the first time to deal with.
Information

Japan Introduces First Artificial Intelligence Law: Promoting Technology R&D Applications and Preventing Abuse

2025-5-28 19:33:16

Information

PetroChina Releases 300 Billion Parameter Kunlun Large Model, AI Intelligence Body Embedded in Gas Station Management System

2025-5-29 11:30:10

Search