ChatGPT outperforms students in exam, raises concerns about AI

A recent study conducted by researchers at the University of Reading in the UK has revealed that AI-generated answers, specifically those from ChatGPT, not only went undetected by exam graders but also received higher grades than those written by actual students.

In a unique experiment, researchers submitted answers generated by ChatGPT on behalf of 33 fictitious students to the examination system of the School of Psychology and Clinical Language Sciences at the same university.

The results were striking: 94 percent of the AI-written responses went unnoticed by the examiners, and about 83 percent of these answers achieved higher scores than the submissions by real students.

The findings, published in the journal PLoS ONE, highlighted the challenge that educators face in distinguishing between human and AI-generated content. "We found that within this system, 100 percent AI-written submissions were virtually undetectable and very consistently gained grades better than real student submissions," the authors noted.

Lead researcher Peter Scarfe, an associate professor at the University of Reading, emphasized the significance of these results. He remarked that the study should serve as a "wakeup call" for educators globally, urging the education sector to adapt to the growing influence of artificial intelligence.

Scarfe pointed out that many educational institutions have moved away from traditional exams to make assessments more inclusive. However, the integrity of these assessments could be compromised by the capabilities of AI.