Building on the first two parts of the collection, this third installment of our special issue series continues the critical examination of how artificial intelligence (AI) is being integrated into the fabric of library and information work. While Part 2 presented a broadening scope of applications and challenges, this collection offers a deeper examination into three specific yet highly significant areas: the cultural dimensions of automated knowledge organization, advanced techniques for extracting structured knowledge from historical texts and the human factors that govern the acceptance of AI-driven public services. Together, these three papers move beyond demonstrating technical feasibility to interrogate the nuanced realities of implementation, accuracy and equity in AI-augmented information ecosystems.The first paper, “Chinese ethnic minority book classification by large language models within CLC,” presents a timely investigation into the promises and pitfalls of applying large language models (LLMs) to a culturally sensitive task. By systematically comparing the performance of DeepSeek-v3 and ChatGPT-4o in classifying bibliographic records of Chinese ethnic minority books, the study does more than benchmark accuracy. It reveals a critical “capability threshold” where models struggle with the fine-grained, culturally embedded knowledge required for accurate subclassification. The high rate of ethnic misclassification and the generation of fabricated codes underscore a vital lesson: technical proficiency in general domains does not guarantee cultural competency. This research serves as an essential warning and a guidepost, arguing powerfully for the development of culturally-aware AI governance frameworks and for classification systems themselves to evolve toward greater inclusivity.Turning to the foundational work of digitizing cultural heritage, the second paper, “Entity Recognition of Ancient Chinese Book based on Semantic Association and Internal Structural Features among Words,” addresses a core challenge in digital humanities. The proposed GARNET framework, which integrates a domain-specific pre-trained model (GujiRoBERTa) with word-pair relation modeling, data augmentation and ensemble learning, achieves state-of-the-art results in named entity recognition on ancient Chinese texts. Its significance lies not only in its technical performance but in its methodological approach. By modeling the relationships between words, the framework better handles the ambiguous entity boundaries and syntactic complexities of historical language. This work provides a robust, scalable model for transforming unstructured classical texts into structured knowledge, enabling new pathways for historical research and cultural preservation.The third paper, “User acceptance of AI chatbot services in public libraries,” shifts the focus squarely to the end-user. Using the unified theory of acceptance and use of technology and information gap theory, this study identifies the key factors driving the intention to use AI chatbots in public libraries. It finds that while performance and effort expectancy are important, social influence and AI literacy are the strongest predictors. Furthermore, it reveals significant disparities in acceptance based on age and education level. These findings are crucial for moving from mere implementation to successful adoption. They provide concrete evidence that marketing and literacy programs, particularly those leveraging word-of-mouth and tailored to different demographic groups, are not ancillary but central to the equitable and effective integration of AI in public service contexts.Collectively, the contributions in this issue reflect a growing maturity in the discourse. We are seeing a necessary turn from asking “can it be done?” to rigorously assessing “how well it is done, for whom, and under what conditions?” Whether it is ensuring cultural fairness in automated cataloging, achieving technical precision in heritage digitization or fostering genuine user acceptance in service design, these papers emphasize that the sustainable application of AI in Library and Information Services hinges on this more critical, context-aware and human-centered approach.The authors used AI tools to polish the language after writing the draft.
Bu et al. (Mon,) studied this question.