Mathematics Pupils at a High School Employed an AI Tutor Based on GPT-4. Their Scores Decreased.
In a recent study conducted by researchers at the Wharton School of The University of Pennsylvania, the effectiveness of AI tutors in educational settings was put to the test. The study, titled 'Generative AI Can Harm Learning', involved nearly a thousand high school students in Turkey who were in grades 9, 10, and 11 during the 2023-24 school year.
The study found that on practice exams, students with access to AI tutors performed differently depending on the type of AI they used. Those with access to GPT Base, which worked similarly to standard versions of ChatGPT, did 48% better than students without access to an AI tutor. However, students using the more advanced GPT Tutor, which was designed to help students find the answers on their own, performed 127% better.
On the actual test itself, the GPT-based students did 17% worse, and the GPT Tutor group performed on average the same as the control group. The researchers believe that AI tutors, particularly those like GPT Base that give away the answers, might hurt learning as students potentially use it as an answer machine.
However, the study also suggests that AI tutors, when designed correctly, can be helpful in certain contexts. GPT Tutor seemed to mitigate the negative impact of an AI tutor on students, though it didn't help them either. The technology is already good at providing answers, but the design of assessment and educational delivery needs to be rethought.
The key to effective AI tutors lies in their design. They should be responsive, adaptive, and transparent, with teacher guidance central to their deployment. AI tutors should promote learning rather than mere productivity by focusing on ethical, personalized, and interactive educational experiences that enhance students' critical thinking, digital literacy, and engagement with content.
Educator-led customization and control, interactive and dialogic learning, personalization with meaningful context, transparency and explainability, focus on formative feedback and skill development, ethical and equitable design, and a human-in-the-loop approach are some of the key design principles that should guide the development of AI tutors.
In practical terms, this means building AI tutors that are not just automated task-completion machines, but tools that encourage active learning, ethical AI interaction, and personalized understanding. The Federal Aviation Agency bans junior pilots from completely depending on autopilot as an example of skills that should be developed without relying on machines. Similarly, in education, the goal is not just the immediate output, but what the student has actually learned.
As AI technology continues to advance, it's essential to remember that it won't change the inherent challenges surrounding AI and learning without proper design. Sungu, a researcher in the field, believes that even though AI might get more advanced, it won't change the fact that critical thinking and problem-solving are vital skills that students should develop without relying on machines.
More research is required specifically around AI in education to make AI tutors more effective. As we move forward, it's crucial to prioritize the development of AI tools that promote learning and critical thinking rather than just productivity.
Read also:
- Shaping production and consumption tendencies via cosmetic certification
- Quantum Leap Career Fair 2025, Organized by SFAN: Focus on Proficiency, Artificial Intelligence, and Employment in the Future
- Uncovering the Fascinating Historical Roots of Christmas Customs
- Calculator for Determining Gravitational Force
