Effectiveness of ChatGPT in Essay Autograding

ChatGPT is an OpenAI-trained large language model based on the GPT-3.5 and GPT-4 architectures. This model has shown potential in the area of autograding, where it may be utilized to offer students instantaneous, computer-generated assessments of their work. Using ChatGPT’s natural language processing capabilities makes it feasible to assess student work with greater nuance and accuracy compared to conventional assessments such as multiple-choice or fill-in-the-blank styles. This research investigates the efficacy of ChatGPT in automated grading, with a specific emphasis on its capacity to comprehensively evaluate written tasks across diverse academic domains. This paper aims to assess the accuracy and reliability of ChatGPT in automated grading and compare its efficacy with conventional grading techniques. The outcomes of our study indicate that ChatGPT demonstrates significant efficiency in automating the grading process, thereby presenting an exciting possibility for enhancing the accuracy and efficiency of grading across diverse educational contexts. In addition, it proves that ChatGPT model 4 accuracy is much better with almost 50% accuracy than model 3.