A recent study published in Scientific Reports reveals that ChatGPT performs on par with or surpasses the average grades of university students across various subjects such as computer science, political studies, engineering, and psychology.
Surprisingly, despite concerns of plagiarism from educators, nearly 75% of surveyed students expressed their willingness to utilize ChatGPT for assignment help.
Researchers Talal Rahwan and Yasir Zaki conducted an experiment at New York University Abu Dhabi, involving 32 faculty members who contributed student submissions for ten assessment questions in 32 different courses. ChatGPT was tasked with generating answers for these questions, which were then compared to student-written responses by three independent graders, unaware of the source. Impressively, ChatGPT outperformed students in 9 out of 32 courses, demonstrating its effectiveness in most areas except mathematics and economics.
Particularly noteworthy was ChatGPT’s exceptional performance in the ‘Introduction to Public Policy’ course, achieving an average grade of 9.56 compared to students’ 4.39.
The study also surveyed 1,601 individuals across Brazil, India, Japan, the US, and the UK, including a minimum of 200 students and 100 educators from each country. Surprisingly, 74% of students showed interest in incorporating ChatGPT into their workflow. In contrast, educators consistently underestimated students’ inclination towards using ChatGPT and 70% of educators viewed its use as a form of plagiarism.
The study additionally evaluated the accuracy of AI text identification tools GPTZero and AI text classifier. These tools incorrectly identified ChatGPT-generated responses as human-written 32% and 49% of the time, respectively.
These findings hold significant implications for the integration of AI tools in educational settings, prompting potential policy adjustments.
Source Scientific Reports