Researchers have made a groundbreaking discovery in using artificial intelligence to grade physics exams. A recent study published in Physics Magazine reveals that language models like ChatGPT can help university educators provide more consistent and transparent grades for introductory-level physics exams.
The study, conducted by Zhongzhou Chen and Tong Wan at the University of Central Florida, tested the effectiveness of ChatGPT in grading multistep physics problems. The results showed that the AI-generated grades agreed with those of human instructors about 70-80% of the time.
What's more, the AI tool was able to provide personalized feedback to students, explaining how their answers addressed each element of the grading scheme. This feedback was rated as good or excellent by expert instructors in over 87% of cases.
The implications of this study are significant, as AI-assisted grading could save time and money while maintaining the same level of grading quality. This could also enable educators to introduce more meaningful assessments and improve learning outcomes for physics students.
