AIGC-Enabled Interdisciplinary Science Measurement

Abstract

[Purpose/Significance] Generative large language models have revolutionized the natural language processing research paradigm, propelling a new trend in artificial intelligence-empowered social science research. They offer fresh perspectives for quantifying the interdisciplinarity and integration of humanities and social sciences from the standpoint of deep semantic features in texts. [Method/Process] This paper employs ChatGPT to perform discipline classification on academic literature in humanities and social sciences. Through small-sample learning, it identifies discipline-specific knowledge entities from model-generated prediction results. These results are then compared and analyzed about the corresponding disciplines of the journals to propose a quantitative research framework for interdisciplinary studies, which includes metrics such as interdisciplinary richness, interdisciplinary closeness, and centrality, alongside interdisciplinary degree. [Results/Conclusion] Focusing on AIGC's empowerment of interdisciplinary science measurement research, this paper introduces a comprehensive research framework and methodology. It addresses issues related to discipline classification, discipline entity extraction from generative model responses, multi-disciplinary candidate question weighting, and content-based metrics for interdisciplinary science. These contributions allow for the thorough utilization of AIGC in social science research and offer valuable insights for exploring the underlying logic of various social science studies.