How AI Vaporizes Long-Term Learning

The use of artificial intelligence (AI) chatbots in the classroom has sparked debate in the education community. Proponents argue that these tools can significantly aid students, while skeptics, including teachers, express concern. A 2024 study on how AI affects learning—involving approximately 1,000 high school students—explored this issue.  

In question was students’ ability to effectively integrate AI assistance into their learning and distinguish between AI assistance and their own understanding. The research tasked students with attending a math lesson and then solving related problems using either traditional methods—like notes and textbooks—or AI tools, including a basic version of ChatGPT and a specially developed AI tutor. Initially, students using AI tools showed remarkable improvements, with those using the basic and tutor versions scoring 48 percent and 127 percent better than their peers, respectively. However, this success was short-lived: On a closed-book test of the same skills, the scores of the students who had used AI plummeted. 

The study revealed that students using the basic ChatGPT engaged in superficial interactions, frequently asking the AI direct questions like “What is the answer?” The researchers concluded that unsupervised use of AI during skill-building phases can become a hindrance—and potentially limit learning.

Despite these challenges, AI’s potential in education should not be dismissed. The software has shown promise when it has been custom-designed to facilitate learning, such as by refusing to provide direct answers and instead prompting students with probing questions. The research on how AI affects learning shows that while the technology is advancing rapidly, it is not yet ready to replace teachers. AI remains limited by the capabilities of its human creators, highlighting the need for careful integration into educational practices to enhance, rather than inhibit, student learning.

To dig more deeply into the research cited in the video, you can find Hamsa and Osbert Bastani, Alp Sungu, Haosen Ge, Özge Kabakci, and Rei Mariman’s 2024 research paper on how generative AI can harm learning. 

To learn more about 2024’s most compelling research findings for educators, visit “The 10 Most Significant Education Studies of 2024.”

Editor’s note: This summary was written with support from Edutopia’s custom AI tool, which prioritizes information from our archive of educational content.