Researchers reported that identifying AI-generated answers could be challenging for exam graders, as their study revealed these responses were not only difficult to detect but also received higher grades compared to those submitted by actual students.

At the University of Reading, UK, researchers submitted answers generated by ChatGPT on behalf of 33 fictitious students to the School of Psychology and Clinical Language Sciences’ examination system. They found that 94% of the AI-generated answers went unnoticed, with approximately 83% of these responses scoring higher than submissions from real students.

“In this study published in PLoS ONE, we observed that AI-generated submissions were indistinguishable within the current assessment framework and consistently achieved superior grades compared to genuine student submissions,” the authors noted.

The researchers emphasized that these findings should alert educators globally to the implications of artificial intelligence in education. Lead researcher Peter Scarfe, an associate professor at the University of Reading, stressed the importance for educational institutions worldwide to adapt to the challenges posed by AI.

“Our research underscores the need for a thorough understanding of how AI will impact the integrity of educational assessments, especially as many institutions move towards more inclusive assessment methods,” Scarfe added.

Source: PTI

Also read: ChatGPT can reshape children’s learning