Artificial Intelligence poses learning risks for students

The uncontrolled use of Large Language Models (LLMs) like ChatGPT in education is raising concerns regarding learning reliability and student development. While these tools can increase efficiency, research indicates significant risks to cognitive growth.

A 30-month study in China involving 26,811 students aged 7–12 revealed that while LLM use increased homework grades by 18% and reduced completion time, it had detrimental effects on actual learning. In classroom settings without AI assistance, grades dropped by 20%, and entrance exam scores for high schools and universities fell by 24% and 18%, respectively. After two years of systematic reliance, many students exhibited new learning problems, particularly among younger students, boys, and high-performing learners.

Experts warn that when mental labor is outsourced to external tools, students fail to build a gradual cognitive foundation. This is especially critical for foundational subjects like syntax and mathematics. Microsoft’s AI lead has cautioned that without sufficient safety safeguards, the development of AI could lead to a new type of silicon-based competitor to humans.