Artificial Intelligence for Science and the Changing Role of University Libraries in China: A Systematic Review of Research Support Services, User Requirements, and Institutional Practice

Abstract

Artificial Intelligence for Science (AI4S), often referred to as the ‘fifth paradigm' of scientific research, is changing the way hypotheses are formulated, experiments are designed, and knowledge is generated in any field. In the era of scientific and technological self-reliance, Chinese universities are also transforming themselves from the traditional role of providing documents to a new role of providing reference services, and their libraries are also changing their functions from document provision to specialized, AI-based Research Support Services (RSS) that are integrated into the research process. This paper reviews and summarizes the recent academic and institutional papers on AI4S and the changing role of Chinese university libraries. This paper summarizes and reviews the academic and institutional literature on AI4S and the changing role of university libraries in China. Through literature review on library and information science journals, science-policy papers from different nations, and case studies of major libraries such as Peking University Library and the Beijing Institute of Technology Library, the review chronicles the development of academic library services in China; explores the user requirements for AI oriented RSS; summarizes the AI-enabled library services currently available, covering intelligent literature retrieval, research data management, AI assisted writing and publishing support, conversational agents, and research evaluation; and reviews key challenges facing libraries, including technical limitations of AI algorithms, ethical and data-property issues for AI, dependency upon AI, and institutional constraints related to funding, training, and strategic alignment. The review also examines the practice in China with reference to international developments and presents a model of competencies and services to help further transform university libraries from a passive service provider to an active partner of scientific discovery. Finally, the paper raises some questions for further research, such as the longitudinal assessment of the effectiveness of RSS, the standardization across institutions, and the governance of generative AI for scholarly workflows.