Implementation and Evaluation of an Automatic Scoring System for Experimental Reports Based on ChatGPT

This paper introduces an automatic scoring system based on ChatGPT for teachers in online laboratories. Students used to be requested to complete experimental reports after the experiments in whether conventional laboratories or online laboratories. However, it is repetitive and tiring work for teachers to mark dozens or even hundreds experimental reports. Moreover, evaluations of students experiment performance are based on teachers’ teaching experience and sometimes are influenced by more subjective factors, so grades marked maybe not objective and fair. Take disadvantages of manual evaluations above into consideration, this paper proposes an automatic scoring system based on ChatGPT where each student’s score is assessed automatically based on his experiment steps and answers to questions in online laboratories. This system is embedded into the Networked Control System Laboratory (NCSLab) and applied into grade evaluations of the Automatic Control Theory experiment course in Wuhan University. Through testing and validation,the system performs the high exactitude and reliability.