COMPARATIVE ANALYSIS OF CHATGPT AND RE3DATA.ORG FOR FINDING DATA REPOSITORIES IN SOCIAL SCIENCE

Abstract

Artificial intelligence (AI) is increasingly important in scholarly communication. Despite concerns about academic integrity compliance, AI tools offer potential benefits for researchers navigating the complex landscape of research data repositories. This study explores whether Chat Generative Pre-training Transformer (ChatGPT) can effectively identify and recommend quantitative and qualitative datasets in social sciences. We compare how ChatGPT (version 3.5) identifies data repositories versus the specialized Re3Data.org registry. The results revealed that ChatGPT can respond with relevant repository recommendations that complement rather than duplicate those found through Re3Data.org, providing researchers with a broader range of options. Standard searches using Re3Data.org offered more structured results with disciplinary categorization, while ChatGPT provided repositories with richer contextual information about their contents. In specialized searches for datasets on generative AI in academic contexts, ChatGPT demonstrated the ability to identify specific datasets across multiple repositories with detailed metadata. However, when asked about broader empirical trends, such as the proportion of quantitative versus qualitative research, ChatGPT could only provide generalized responses without precise statistics, highlighting its limitations in accessing current empirical data. The conclusion reached is that while ChatGPT cannot yet generate repository data of suitable quality for advanced-level analyses in all contexts, it is a valuable complementary tool to traditional repository registries. As AI tools continue to develop, educators and scholars must shift their focus from negative expectations to the practical benefits these tools can provide in research data discovery.