According to TapTechNews on June 27, college students have begun to use AI to complete their schoolwork, and the average score of the answers submitted by AI is higher than that of real people, and it is difficult to detect.
Peter Scarfe, an associate professor at the School of Psychology and Clinical Language Sciences, University of Reading, UK, said, College students have begun to use AI tools such as ChatGPT to complete their schoolwork, not only can they get higher scores, but it is also difficult to judge.
TapTechNews translated Professor Scarfe's interview content: The data in our research indicates that it is very difficult to detect AI-generated answers. There has already been a lot of discussion about AI. Completing schoolwork for college students on behalf of them is also a form of AI, but the sample size of this survey is relatively limited.
In the study published in the Plos One magazine, Professor Scarfe and his team used GPT-4 to generate answers to exam questions and submitted these answers on behalf of 33 fake students.
When the graders were unaware, they evaluated the scores and found that the answers to the undergraduate psychology schoolwork generated by AI were not detected in 94% of the cases, and on average, the score was higher than the score actually submitted by the students.