The first phase of the Reimagining Discovery project at Harvard Library sought to address the challenge of fragmented search experiences of special collections materials using artificial intelligence (AI) technologies, such as embedding models and large language models (LLMs). The resulting platform, Collections Explorer, simplifies and enhances the search experience for more effective special collections discovery. The project team took a user-centered and trustworthy approach to implementing AI, grounding the choices of the platform in user empowerment and librarian expertise. The development process included extensive user research, including interviews, usability testing, and prototype evaluations, to understand and address user needs. Collections Explorer was developed using a multi-component architecture that integrates multiple types of AI. The team evaluated more than 12 models to select ones that were the best fit for the need, as well as being ethical and sustainable. Detailed system prompts were developed to guide LLM outputs and ensure the reliability of information. The methodical and iterative approach helped to create a flexible and scalable platform that could evolve to support other material types in the future. Initial research showed that potential users are enthused at the prospect of AI-powered features to enhance discovery, especially the item-level summaries and related search suggestions. The project demonstrated the potential of integrating AI technologies into library discovery systems while maintaining a commitment to trustworthiness and user-centered design.
Caizzi et al. (Mon,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: