AI Helps Students Score Better In Homework But Hurts Exam Results, Study Finds

Artificial intelligence (AI) tools are making it easier for students to complete homework, but better assignment scores may not always mean better learning. A new study of students in China has found that while the use of generative AI improved homework performance, exam scores declined over time.

The research, titled The Generative AI Learning Penalty: Evidence from Chinese Secondary Education, examined data from 26,811 students in Grades 7 to 12. It looked at students’ homework scores, the time they spent completing assignments, monthly examination results and performance in high-stakes entrance examinations.

The findings raise questions about how students use AI tools such as ChatGPT for schoolwork and whether depending on these tools can affect their ability to answer questions independently.

Homework Scores Rise, but Exam Performance Declines

According to the study, adopting generative AI was linked to an 18% increase in homework scores and a 30% reduction in the time spent completing assignments.

However, the improvement in homework performance did not translate into better results in classroom examinations. Within six months of first using AI, students’ monthly in-class exam scores fell by 20%.

The researchers also found that students who had used AI scored lower in senior high school and college entrance examinations. Their scores were, on average, 5% to 7% lower than those of students who had never used the technology, according to the reported findings.

The results suggest that completing assignments more quickly and receiving higher homework marks may not necessarily reflect a deeper understanding of the subject.

Why Depending on AI for Homework Can Be a Concern

One possible explanation is that students may use AI to produce answers without spending enough time understanding the concepts behind them. While this approach can help complete assignments, it may leave students less prepared to solve unfamiliar questions during examinations.

The study found that learning losses were concentrated among students whose behaviour suggested they were outsourcing their homework to AI. These students received high homework scores despite spending considerably less time on assignments.

By contrast, students who used AI but spent a similar amount of time on their homework as students who did not use the technology experienced much smaller learning losses.

This distinction indicates that the impact of AI may depend less on whether students use it and more on whether they use it to support learning or replace their own effort.

What the Findings Mean for Schools and Students

The findings highlight the need to encourage responsible use of AI in education. Instead of relying on AI-generated answers, students can use these tools to clarify difficult concepts, seek feedback and practise solving problems.

Teachers may also need to place greater emphasis on supervised classroom assignments, discussions and examinations that help assess students’ independent understanding.

The study does not suggest that AI has no place in education. Rather, it highlights the risk of using the technology as a shortcut to completing academic work without learning the material.

As AI becomes more common in classrooms, the challenge for schools will be to help students benefit from the technology while continuing to develop their own reasoning, problem-solving and subject knowledge.

Also read:

AI Impact On Jobs: Goldman Sachs Warns Middle Management Roles Could Be Reshaped

4 Indian B-Schools Rank Among Global Leaders In Teaching Cases, IIM Bangalore Leads

Stay updated in real-time!
📢 Join our WhatsApp Channel: https://whatsapp.com/channel/0029Vb2Mqc8GJP8L2bwzC10z
🔔 Join our Telegram Channel: http://t.me/jagranjoshofficial