AI Advances in Exams, Sparking Plagiarism Concerns
Artificial Intelligence (AI) has surpassed human students in exams, raising concerns about increased plagiarism in the education sector, according to a recent study published in Scientific Reports. Researchers from New York University Abu Dhabi delved into the implications of ChatGPT, an AI model that utilizes generative pre-trained transformers (GPT) in natural language processing.
The GPT model’s capabilities lie in its ability to generate coherent and relevant texts based on contextual understanding. Trained on a vast amount of online text, it has mastered statistical patterns, grammar, and linguistic context. GPT models, including ChatGPT, have gained popularity in various applications such as chatbots, virtual assistants, and text completion tools.
For the study, ChatGPT was tested against university students in a range of subjects, including computer science, political science, engineering, and psychology. Researchers gathered questions from faculty members teaching various courses at NYUAD. ChatGPT was asked to generate three distinct answers for each question, which were then compared with responses from three students.
The results showed that in nine out of 32 courses, ChatGPT received comparable or superior grades to the students. However, students consistently outperformed the AI model in mathematics and economics. Notably, in an introduction to public policy course, ChatGPT significantly outperformed students, receiving an average grade of 9.56 compared to the students’ 4.39.
Despite educators dismissing the use of AI tools like ChatGPT as plagiarism, the study revealed that 74% of students surveyed indicated they would use ChatGPT for their assignments, while 70% of educators considered such use as plagiarism. This disparity highlights the tension between students’ desire to utilize AI tools surreptitiously and educators’ efforts to detect and prevent plagiarism.
The study also raised concerns about the ownership of data used to train AI models and the difficulty in distinguishing between human and AI-generated content. Schools in New York City, Los Angeles, and Baltimore have already banned the use of ChatGPT, and Australian universities are returning to handwritten exams to combat AI-generated essays.
To address the issue, some educators have phased out take-home assignments, while conferences and journals have prohibited the use of ChatGPT for academic writing. However, proponents argue that AI tools like ChatGPT can enhance students’ writing skills, particularly for non-native English speakers.
Overall, the study highlights the challenge that educational institutions face in developing academic integrity policies that account for the extensive applications of generative AI. As AI continues to advance, it is essential to consider its impact on students’ learning journey and their ability to express ideas in their own words.
