A recent study highlighted by the World Bank raises important concerns about the impact of generative artificial intelligence on education and human capital development. While AI is increasingly being explored as a tool to improve education, healthcare, and other public services, researchers are beginning to examine the unintended consequences of widespread and uncontrolled AI use. The findings suggest that the effects of everyday AI usage may be far more significant than those observed in carefully designed educational interventions.
The study followed students in grades 7 to 12 in a county in China over a period of two and a half years beginning in 2022. Researchers compared students who independently adopted AI tools with those who did not, using differences in adoption timing to estimate the effects of AI use on academic performance. The results revealed a striking contrast between short-term task completion and actual learning outcomes.
Students using AI spent approximately 30 percent less time on homework while achieving homework scores that were about 18 percent higher than before. However, despite these apparent gains, their academic learning outcomes deteriorated significantly. Monthly examination scores dropped by around 20 percent after five months of AI use, while performance in major entrance examinations declined by 18 percent and 24 percent. The negative effects were observed across subjects, with the largest impacts seen in social sciences, followed by science, technology, engineering, mathematics, and language-related subjects.
The findings suggest that many students may be relying on AI to complete assignments rather than engaging in the cognitive effort necessary for learning. By providing ready-made answers and solutions, AI tools can reduce the need for critical thinking, problem-solving, and independent analysis. As a result, students may achieve better homework results while gaining less knowledge and understanding of the subject matter.
Researchers note that the size of the observed decline is exceptionally large when compared with other education interventions. The negative impact on exam performance substantially exceeds the learning improvements reported in many AI-assisted tutoring programs. While the study acknowledges that other factors could partially influence the results, the authors argue that such factors are unlikely to fully explain the dramatic decline in academic performance.
The implications extend beyond education. As AI becomes more integrated into daily life, people increasingly use it for health information, career guidance, skills development, and emotional support. These uses often occur outside formal systems where impacts can be carefully measured and evaluated. Consequently, the broader effects of AI on human development may be shaped more by how individuals choose to use the technology than by structured programs designed by governments or institutions.
The World Bank emphasizes that more research is needed across different countries and educational settings to determine whether similar patterns emerge elsewhere. Understanding the real-world impact of AI on learning, skills acquisition, and human capital development will be essential for ensuring that the benefits of artificial intelligence are realized while minimizing potential risks to future generations.
