关键词: AI素养教育, 生成式人工智能, 高校图书馆, 香港, 教育实践, 信息素养

Abstract:

[Purpose/Significance] The rapid advancement of generative artificial intelligence has reshaped higher education, bringing both opportunities and challenges such as academic integrity concerns, algorithmic bias, and data privacy issues. Consequently, AI literacy has emerged as a global educational priority. University libraries, as traditional hubs of information literacy, are undergoing a transformation from resource providers to literacy enablers, with international library organizations explicitly calling for libraries to play a central role in fostering AI literacy. While existing research has primarily focused on conceptual frameworks and macro-level strategies, there remains a notable gap in empirical studies examining the micro-level implementation mechanisms, practical challenges, and cross-departmental collaboration strategies of university libraries in delivering AI literacy education. This study addresses this gap by investigating the practices of university libraries in Hong Kong, a region with a distinctive "policy-practice" collaborative model, aiming to provide actionable insights for university libraries in China's mainland. [Method/Process] This study employed an online research method combined with content analysis. Between December 2025 and March 2026, the official websites of 14 university libraries in Hong Kong were systematically accessed to collect publicly available information, covering diverse institution types, including comprehensive, polytechnic, liberal arts, and arts universities. Three types of data were collected: AI policies or guidelines at both institutional and library levels; AI literacy resource navigation structures, particularly the organization of LibGuides and educational practice activities, including workshops, training courses, online modules, and thematic lectures. Descriptive research methods were used to categorize the collected information, while comparative analysis was conducted to identify commonalities and differences across the 14 libraries. Additionally, a quantitative scoring method (0 to 3 scale) was applied to evaluate the coverage depth of various AI literacy education themes, enabling the generation of a thematic distribution heatmap. [Results/Conclusions] The findings reveal that university libraries in Hong Kong have established a three-pronged AI literacy education model characterized by "policy guidance + resource integration + diversified teaching." A dual-layered policy ecosystem has emerged, wherein institutional policies set ethical boundaries and libraries translate these policies into actionable educational practices. Libraries have developed systematic knowledge portals via LibGuides, forming complete knowledge chains encompassing foundational concepts, tool applications, ethical norms, and academic support. Educational activities take multiple forms, including thematic workshops, embedded instruction within disciplinary courses, systematic online courses, and targeted lectures. Content design equally emphasizes both technical skills and ethical reasoning, with some libraries developing operational frameworks for critically evaluating AI-generated content. However, several challenges persist: insufficient policy independence at the library level, uneven distribution of educational content, a lack of comprehensive evaluation mechanisms, limited integration with disciplinary curricula, and bottlenecks in librarian professional development. Compared with existing studies, this research reveals the underlying "policy-practice" synergy that enables the translation of institutional policies into concrete educational actions, contrasting with the fragmented state of AI literacy education observed in some mainland Chinese universities. The Hong Kong model represents an intermediate form between "policy absence" and "deep embedding", offering a phased and referable pathway for mainland Chinese university libraries. The study has two limitations: reliance on publicly available online information and the lack of longitudinal data. Future research should incorporate surveys, interviews, and longitudinal assessments to develop robust evaluation frameworks.

Key words: AI literacy education, generative AI, university library, Hong Kong, educational practice, information literacy