BEIJING: Chinese academic Xia Wenbin urged universities to pursue AI education reform by prioritising human judgment and critical thinking as artificial intelligence increasingly takes on classroom tasks.
Xia said universities should move from memorisation-based instruction towards problem-solving, evidence verification, ethical reasoning and cross-cultural communication.
Writing in China Daily on July 24, he proposed reorganising curricula around real-world challenges. Courses could examine algorithmic accountability, AI-generated misinformation and technology-related risks.
Xia also called for process-based assessments that examine how students formulate questions and use AI tools. Learning and AI-usage logs could better support academic integrity than automated detection software alone, he argued.
Teachers should serve as editors, mentors and ethical advisers, Xia said. Meanwhile, universities should expand learning through community projects, internships and international collaboration.
He identified AI hallucinations as a major educational challenge. Students need sufficient subject knowledge to trace sources, compare evidence and detect plausible but inaccurate machine-generated answers, he wrote.
China’s Ministry of Education and eight other departments issued a digital-education guideline in 2025. It called for AI integration across teaching, learning, assessment and research while strengthening critical thinking and practical skills, according to an official government policy report.
Xia cited existing applications at Tsinghua University, Northwest Agriculture and Forestry University, Huaqiao University and Xi’an Jiaotong-Liverpool University. These included collaborative learning, calculus visualisation, digital-video development and AI-assisted translator training.
Xia is dean of the School of Marxism at the Southern University of Science and Technology and a former president of the University of International Business and Economics. China Daily identified the article as commentary and said his views did not necessarily represent the publication.
