I'm interested in know how well ChatGPT or other models of LSTMs are capable of grading a code basing on it's merits of how well designed and well coded are like a human programming teacher will. Grade a answer of a student of introduction to programming for instance. Today, we can measure time complexity and syntax, also if the algorithm produced pass or not given a set of tests. This is used to auto corrrect/feedback the student to give a better answer. But in a situation of a final exam, when feedback is given after the grading (lets assume A+, A, B, C, D, and E for fail). How well ChatGPT could grade as well as a human? In other more complex words, how positive is the correlation between a grading made by a human and a grade made by an AI?
ChatGPT is an AI language model that can understand and generate natural language text. While it is capable of understanding code and technical language, it is not specifically designed for grading code answers in exams. Its ability to grade code answers would depend on the quality of the criteria provided for grading and the accuracy of the code in answering the given question.
However, grading code answers goes beyond simply checking for correct syntax or output. It often requires an understanding of the logic and problem-solving skills used by the student. These are higher-level skills that require human judgment and cannot be solely assessed by an AI model like ChatGPT. Therefore, while ChatGPT can assist in grading code answers, it is unlikely to be able to provide the same quality of grading as a human professor, especially when it comes to evaluating the reasoning and thought process behind the code.
Ahshanul Haque somebody already tested with a real dataset and saw the results? Although unlikely.
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Programming Languages
- Python