This paper explores the technical defects of the artificial intelligence-generated content (AIGC) paper detection system and its root causes and proposes the technical response strategies from the practical level. Using the AIGC detection service system of CNKI and the AI detector of "Jianziyuan",this research detects 10 legal paper abstracts generated by ChatGPT and 10 similar abstracts rewritten by an AI rewriting software to verify their AIGC identification abilities. Based on the analysis of the detection results,it can be concluded that the AIGC paper detection system has three major technical defects:the low accuracy,the high difference rate and the weak sensitivity. The root causes of the defects are insufficient model training,improper algorithm optimization,and lack of preset lexicons. The study shows that the improvement of the detection system depends on the benign interaction between system research and journal development:at the R&D coordination level,academic journals should communicate and cooperate with technology providers to provide diverse test samples and preset lexicons;at the industry development level,as the ultimate goal,academic journals should avoid the academic misconduct caused by AI ghostwriting and dialectically utilize the high-quality and authentic knowledge contents generated by AI;at the editorial practice level,journal editors should play the roles of human-computer relationship collaborators and guide the optimal development of the detection system.
AI-generated content/artificial intelligence generated content (AIGC)/AI text detection/AIGC detection service system of CNKI/AI director of "Jianziyuan"/technical defect/academic journal/response strategy
维普
万方数据