Panel : AI Transformation in Metadata Research and Practice

This introduction will bring an overview of a recent activity and outcomes of the exploration of global and regional perspectives regarding AI and the transformation of metadata research and practices through the DCMI Education Committee and our professional networks globally since the situations are much more complicated beyond machine-readable and understandable basic languages and resources. A significant project is a large-scale survey (to be reported in this panel first) initiated by our DCMI Education Committee AI Task Group and activated by the committee members with multilingual and global participants. Based on this effort, we also have proactively engaged with our professional networks to gather use cases and solicit chapter contributions, which leads to an upcoming book with the following focused themes: AI Applications for Digitized Resources and Languages; AI for Automating Metadata Processes; Impact and Ethical Considerations of AI in Metadata; and Regional Perspectives on AI Adoption and Professional Development in Metadata. These outcomes will be introduced, demonstrating a global analysis of AI’s current and future influence on metadata including the key trends, anticipated impacts, and perceived benefits and challenges of AI implementation.

This presentation explores the opportunities and challenges of using AI for metadata creation and management, based on findings from "The Survey on Metadata and AI" by the Metadata and AI task group of the DCMI Education Committee, conducted recently, with the responses from 752 library and information professionals from the world. The findings indicate the necessity for continuous professional development for metadata professionals and highlight the importance of evaluating the effectiveness of AI tools and human intervention in metadata creation processes to ensure transparency and reliability. It features the necessity to identify and mitigate biases in AI-generated metadata through controlled vocabularies and community review mechanisms. Finally, it calls for establishing best practices for integrating AI and human workflows, with clearly defined roles and responsibilities. The presentation advocates for a human-centered approach to AI implementation in metadata creation and management. This approach emphasizes the integration of AI tools while retaining human oversight for tasks such as ensuring the accuracy, relevance, and cultural sensitivity of AI-generated metadata, addressing potential biases and ethical implications, and maintaining the quality and relevance of metadata standards for end-users.

This presentation explores the integration of ethics and cultural heritage into the AI development pipeline, proposing a comprehensive framework for evaluating AI performance. It challenges traditional views that assess AI systems solely on technical metrics, arguing that ethical considerations—such as fairness, inclusivity, and cultural sensitivity—must be foundational to defining what it means for an AI system to perform well. The first part of the presentation examines ethics across the AI development pipeline, from data collection and algorithm design to deployment and impact assessment, highlighting the importance of incorporating diverse perspectives at every stage. The second part proposes a shift in how AI performance is evaluated: instead of relying on traditional success metrics, the presentation asserts that AI must meet rigorous ethical standards to be considered truly successful. Finally, the presentation emphasizes the role of cultural heritage principles in shaping AI training data management. By applying archiving practices rooted in cultural preservation, we can create AI systems that respect diverse histories and values, ensuring fairness and inclusivity in training datasets. Ultimately, this approach redefines AI performance to reflect not just technical achievement, but also ethical integrity and cultural respect.

This presentation introduces a prototype workflow that integrates Generative AI with the CRMinf argumentation ontology to support structured metadata generation from scholarly texts in historical research. Using Chinese wooden slip reconstruction as a case study, it demonstrates how interpretive reasoning can be translated into machine-readable records. The presentation also discusses challenges in automating argumentation mapping and examines implications for metadata transparency, provenance, and reuse in cultural heritage and digital humanities contexts.

This presentation reports a study which investigates the potential for AI implementation in Kenya's GLAM (Galleries, Libraries, Archives, and Museums) ecosystem, building on previous research on AI-preparedness in Kenyan cultural heritage institutions. It examines the information management scenarios within these institutions to identify areas that could benefit from AI tools and automation. The research utilizes the DCMI Education Committee AI Task Group’s curated list of AI tools and data from a global AI survey and explores additional AI technologies to match them with specific needs in Kenya's cultural heritage sector. By analyzing these resources and potential applications, the study aims to provide a comprehensive overview of how AI can enhance information management practices in Kenya's GLAM institutions, contributing to their digital transformation and improving their services.