ChatGPT in the classroom: promises and precautions
Since its release in 2022, ChatGPT has sparked intense debate about its role in education. Is it an ally in building knowledge, or a shortcut that undermines cognitive effort? In a recent article published in Computers & Education, Deng and colleagues present the first systematic review and meta-analysis of ChatGPT-based educational interventions. Their conclusion is clear: the tool can improve academic performance and foster motivation, but caution is warranted due to methodological limitations and the novelty effect.
The analysis of 69 studies shows that interventions with ChatGPT are mostly concentrated at the university level, with a marked focus on language learning. In these contexts, students not only show positive attitudes toward the tool but also improvements in performance and self-efficacy. Moreover, evidence suggests that ChatGPT can reduce cognitive load, making it easier to acquire complex knowledge.
Yet enthusiasm must be tempered with prudence. The authors highlight a widespread lack of power analyses, which compromises the reliability of reported effect sizes. They also note that in many experiments, participants were allowed to use ChatGPT during post-tests, making it difficult to determine whether gains were due to genuine learning or simply to the quality of generated answers.
Beyond the immediate findings, the study raises crucial questions: Can the positive effects be sustained over time, once the initial fascination wears off? How can ChatGPT be integrated into pedagogical designs that promote critical thinking and originality, rather than technological dependency? And, perhaps most urgently, what happens in earlier educational stages, which remain largely unexplored?
The contribution of Deng et al. is both a map and a warning. It shows that ChatGPT can be a catalyst for new forms of learning, but also that research needs stronger theoretical frameworks and more rigorous experimental designs. The task now is twofold: to harness the opportunities of generative AI while setting boundaries and conditions for its transformative impact on education.
Reference
Deng, R., Ben Ali, A., Hämäläinen, R., Crompton, H., Chen, L., & Yang, X. (2025). ChatGPT in education: A systematic review and meta-analysis. Computers & Education, 212, 105265.
https://doi.org/10.1016/j.compedu.2024.105265