Detection of content generated by artificial intelligence in education

Artificial Intelligence (AI) language models, like ChatGPT, have recently had a great impact on society. These AI tools have advanced capabilities and present a unique opportunity for students to potentially, and dishonestly, use them for assignments. Thus, academic researchers and teachers must now investigate the ways to detect whether a text was generated by AI or not. For this purpose, there are detection tools on the Internet based on machine learning algorithms and techniques of natural language processing to identify features and patterns of AI-generated content. However, these detection tools may produce false positive results and innocent students may be wrongly accused. This paper first discusses the features of AI-generated content and explains the logic of detecting it automatically with the aid of software tools. Then, several detection tools currently available on the Internet are described. The paper also presents the results of a survey that collected data about teachers’ attitudes to cheating behaviors enabled by technologies and the relevant issues. The survey results show that teachers are concerned about the usage of AI in cheating and the accuracy of the software tools detecting AI-generated content. In conclusion, the paper provides some suggestions to alleviate the problems.