ScholarWorks@중앙대학교

초록

Since the introduction of ChatGPT in November 2022, generative AI (GAI) services have been intensively studied in the field of AI education. This is because GAI services can provide more human-like interactions and more plausible responses compared to traditional AI services. However, GAI services can generate responses that include AI hallucinations due to factors such as training data bias and model bias. In relation to this subject, some studies investigate the causes of AI hallucinations and explore strategies to mitigate their negative impact. However, the number of domestic studies remains smaller than that of international studies. In this study, we analyze international research trends related to AI hallucinations and review research cases to understand the current status of domestic AI hallucination research. Based on our findings, we propose directions for the educational use of GAI that account for the issue of AI hallucinations. Our analysis highlights the need for both a clear understanding of AI hallucinations and the practical application of methods to mitigate negative impacts for the educational use of GAI.

키워드