Today, universities worldwide are actively embracing generative artificial intelligence (GenAI) tools, exemplified by ChatGPT. A 2024 global survey revealed that a staggering 86% of university students frequently use AI in their daily studies.

The rapid adoption of AI in education has also sparked widespread controversy. A report on the website of the British journal *Nature* states that many believe AI can help improve the quality of education and help students cope with a rapidly changing world; however, many others worry that over-reliance on AI might inhibit the development of students' independent thinking and critical thinking skills.

AI is sweeping university campuses.

The high rate of AI use among teenagers is remarkable. This year, the artificial intelligence company Anthropic analyzed 1 million anonymous conversations between college students and their generative AI tool "Claude," finding that students majoring in science, technology, engineering, and mathematics (STEM) used it significantly more than those majoring in business and humanities.

What are the main purposes for students using these generative AIs? A survey of over a thousand British students conducted by the Higher Education Policy Institute, a think tank at the University of Oxford, shows that most students use AI to write, optimize, or summarize texts.

Meanwhile, tech giants like OpenAI and Google are actively promoting their products to universities and students. Many universities are also embracing this trend. A team at the University of Sydney in Australia developed the GenAI platform Cogniti based on the ChatGPT interface. Since its launch in 2023, the platform has been embedded in the university's digital teaching system, allowing teachers to customize AI assistants for different courses—serving as intelligent tutors for science modules and transforming brief grading comments into detailed academic feedback. Currently, over a thousand teachers at the university actively use the platform, and its results are shared with more than a hundred universities worldwide.

Last April, scientists at Tsinghua University constructed a three-layered AI teaching integration architecture: the bottom layer integrates AI models from multiple companies, including DeepSeek, Alibaba Cloud, OpenAI, and Google; the middle layer features a "knowledge engine" to provide accurate and cutting-edge information for various disciplines; and the top layer integrates various student platforms, such as teaching systems equipped with AI teaching assistants and freshman guidance robots. It is reported that this system has been adopted by hundreds of universities in China.

The impact on academic performance is mixed.

Do students using AI tools truly gain more knowledge? Preliminary research results have raised warning signs. A study from Tsinghua University shows that students using AI learning companions initially outperformed non-users in after-class tests, but this advantage reversed after two to three weeks, with AI users scoring lower than the control group.

For the first time, MIT's Media Lab has revealed the impact of AI-assisted writing on brain activity through neuroscience research. Their study, published at the end of June, showed that students who relied on ChatGPT to complete writing tasks experienced significantly reduced brain activity. This research has resonated widely within the global education community—researchers have received 4,000 emails from primary, secondary, and university teachers expressing concerns about students' over-reliance on AI.

Although the study had a limited sample size and did not consider long-term effects, it did reveal a common concern: students who rely too heavily on AI may be losing critical thinking skills—the core competencies of deeply analyzing information and making rigorous judgments. Anthropic's analysis corroborates this concern: in the millions of student-AI conversations it analyzed, nearly half of the students directly requested answers or ready-made content.

Olivia Geist, a professor at Radboud University Radboud in Radboud, Netherlands, also stated that she has observed a significant decline in students' abilities in recent years, with some students finding it difficult to independently complete papers or in-depth articles. In light of this, in June of this year, more than 1,000 experts worldwide published an open letter opposing the uncritical adoption of AI technology by universities.

It is encouraging that some students have begun to consciously reduce their use of AI, realizing that over-reliance will hinder their intellectual development.

The controversy remains

However, Ravi Bellamkunda, provost of Ohio State University, believes these concerns may be exaggerated. He uses the analogy of a calculator replacing basic calculations, suggesting that delegating certain tasks to AI can free up higher-order thinking space for the brain.

Some studies have shown that GenAI tools can help students learn. An experiment by Greg Kostin's team at Harvard University showed that undergraduates receiving personalized AI tutoring in physics experienced a significant improvement in learning efficiency. Among nearly 200 students participating in the experiment, the median test score in the AI-tutored group jumped from a baseline of 2.75 (out of 5) to 4.5, far exceeding the 3.5 score in the traditional teaching group.

Kostin believes the key lies in how it's used; AI can both replace thinking and lead to complacency, and it can also be a powerful tool for promoting deep thinking. The core of education lies in using AI to truly stimulate students' critical thinking and enthusiasm for learning.

Shafka Isaac, head of the Department of Educational Technology and Artificial Intelligence at UNESCO, cautioned that as teachers and students rapidly adopt various GenAI tools, relevant policies, teaching methods, and ethical guidelines urgently need to keep pace.