IA et éducation : le piège des devoirs faits par ChatGPT, par Franck Ramus

A growing body of research suggests that students using AI tools such as ChatGPT to complete homework may be undermining their ability to learn effectively. A recent study conducted on nearly 27,000 secondary school students in China has raised concerns about how generative artificial intelligence impacts academic performance. The findings, published in June by three researchers, David Strömberg from the University of Stockholm and Victor Lei and Yanhui Wu from the University of Hong Kong, suggest that heavy reliance on AI for assignments could lead to diminished learning outcomes. The study analyzed data spanning 30 months, covering students equivalent to grades five through final year of high school. It focused on how AI-generated content affects students’ capacity to acquire knowledge and develop critical thinking skills. The Chinese education system proved ideal for this kind of research due to its widespread use of digital platforms for homework, monthly exams, and national assessments. During the study period, from January 2023 to June 2025, approximately 80 percent of participants began using AI models such as Doubao or DeepSeek regularly for their schoolwork. According to the researchers, the results were troubling. Students who frequently used AI to complete tasks showed lower levels of comprehension and retention compared to those who did not rely on such tools. The study also found that these students struggled with more complex problem-solving and exhibited reduced motivation to engage deeply with their coursework. While the study does not claim that AI usage is entirely detrimental, it highlights the risks associated with over-reliance on automated solutions. The implications of the study have sparked discussions among educators and policymakers in France and other countries facing similar challenges. In France, there has been increasing concern about the role of AI in student learning, particularly following recent statements by government officials advocating for greater integration of AI into the curriculum. One proposal under consideration involves dedicating one hour per week to teaching students how to use AI responsibly and ethically. In addition to policy debates, the issue has gained traction in political circles. Candidates running for president in upcoming elections have proposed various measures aimed at regulating AI use in schools. Some suggest implementing stricter guidelines on AI access, while others advocate for increased teacher training to help students navigate the ethical and practical aspects of AI-driven learning. French media outlets, including L'Express and Le Figaro, have covered the study extensively, highlighting both its significance and the broader implications for educational practices. L'Express noted that the findings align with global trends showing a rise in AI adoption within classrooms, raising questions about whether traditional methods of instruction are being sidelined. Meanwhile, Le Figaro emphasized the need for a balanced approach, suggesting that AI should complement rather than replace human-led learning. As the debate continues, experts warn that without proper oversight, the use of AI in education could lead to long-term consequences for students’ cognitive development. Researchers stress the importance of fostering independent thinking and ensuring that technology serves as a tool rather than a crutch. With governments and educators increasingly looking to AI as part of modern pedagogy, the challenge lies in finding ways to harness its benefits without compromising fundamental learning principles.