Abstract

The emergence of ChatGPT, a generative, pre-trained transformer, has garnered significant attention in various fields, including education. To better understand the implications of this new tool, we sought to investigate the perspectives of practitioners and researchers on its use. A systematic review of the existing literature was conducted using the Web of Science database, and a set of filter criteria was developed to guide the search. We then applied these criteria to the search results for the two years following ChatGPT’s appearance, resulting in 65 papers for the first round and 36 papers for the second round. To analyze the 101 selected papers, three computational tools were employed: Latent Dirichlet Allocation (LDA) topic modeling, standardized ChatGPT-based coding, and Python for sentiment analysis. The results indicate that: 1) Early discourse is more concentrated around integrity/reliability concerns, while later discourse becomes more practice-oriented; 2) Researchers generally expressed neutral to positive sentiments towards ChatGPT in both time periods, with a more pronounced positive trend emerging in the second round analysis; 3) Disagreement has not disappeared; rather, it appears in narrower concern categories and more specific implementation scenarios. We would like to recommend: First, institutions need explicit and field-sensitive guidance on acceptable GenAI use, especially for assessment and academic-integrity boundaries. Second, because concerns vary by role (students, instructors, professional programs), governance should be role-specific rather than uniform. Third, evaluation systems should move beyond detection-only responses toward transparent task design, process-based assessment, and accountable AI-use disclosure.

This work is licensed under a Creative Commons Attribution 4.0 License.