A new study on the impact of artificial intelligence tools on student learning has sparked debate online after Zoho founder Sridhar Vembu shared a warning that overdependence on AI in classrooms.

Commenting on a social media post discussing the research, Vembo stated on X, “This research shows that high school students don’t learn mathematics with AI, they just learn to depend too much AI. AI is not a training wheel, it becomes a crutch.”

The study, underlined in a detailed thread by user Rimsha Bhardwaj, was conducted by researchers led by Hamsa Bastani, a professor at the Wharton School of the University of Pennsylvania. As per the post, around 1,000 high school students in Turkey participated in the experiment and were examined to determine whether AI tools actually enhance learning outcomes in mathematics.

ALSO READ

What did the study test?

The research reportedly divided students into three groups during four separate 90-minute math practice sessions.

One group used a regular GPT-4 chatbot that was capable of directly answering questions. Another used a modified “GPT Tutor” version designed to give hints and guidance instead of a final solution. A third group studied by using only textbooks and independent problem solving.

According to the post, students using AI initially appeared much better during the practice period. The GPT Base group reportedly solved 48% more problems in comparison to students who were working without AI, whereas the GPT Tutor group solved 127% more.

Charts shared along with the post also indicated that students using AI exchanged more messages per problem and had longer discussions with the chatbot compared to those who were solving problems without assistance.

What happened when AI was removed?

The final result indicated a very different outcome.

According to the post, students who practised with the standard GPT interface scored 17% lower in the actual examination in contrast with students who studied without AI tools.

Researchers reportedly concluded that many students were relying on AI-generated answers instead of developing own-problem-solving skills.

ALSO READ

The thread claimed that analysis of chat logs demonstrated that several students quickly taking out answers from the AI system rather than taking the time to understand the logic behind the problems.

One line from the paper quoted in the thread mentioned that “without guardrails, students attempt to use GPT-4 as a crutch during practice, and subsequently perform worse on their own.”

“The conversation around AI in education must now evolve from ‘how much AI’ to ‘how responsibly AI is integrated.’ Technology works best when it augments human learning rather than shortcuts it. Especially in subjects like mathematics, where logical progression and analytical discipline are foundational, AI should function more like a guided mentor than an automated answer engine,” Dr Aloknath De, Founder and CEO at TechCrafter said.

Why is the study drawing attention?

The discussion has gained attention because it raises broader questions about how AI tools should be used in education.

As per the survey findings, students who relied heavily on AI were often more confident about their knowledge of the subject, despite not performing well in tests conducted without AI assistance.

“The crutch had made them confident and weak at the same time,” the post stated while explaining the study’s findings.

Vembu’s remarks added to the online debate around whether AI tools should act as assistants for learning purposes or whether unrestricted use may hamper critical thinking and independent reasoning skills among the students.

Disclaimer: This story is based on a post shared by a social media user. The details, opinions, and statements quoted herein belong solely to the original poster and do not reflect the views of Financialexpress.com. We have not independently verified the claims.`