Yuhong Yuan
Yuhong Yuan, PhD, is a former gastroenterologist, former Cochrane information specialist, and former Cochrane managing editor. Dr. Yuan is currently an adjunct professor at Western University, a part-time assistant professor at McMaster University, and a senior research associate at London Health Sciences Centre. She has been involved in numerous research projects and has contributed extensively to clinical guideline development as a health research methodologist and guideline methodologist. Dr. Yuan has co-authored over 280 peer-reviewed medical publications and more than 190 conference proceedings, with an h-index of 74 on Google Scholar.
Sessions
AI agents and agentic AI systems represent a new architectural paradigm for applying Generative AI in workflows that require reasoning, information retrieval, and decision support. Unlike earlier forms of GenAI that focused primarily on text generation, agentic systems integrate retrieval, ranking, planning, and tool-use to deliver more reliable, verifiable, and domain-specific outputs. Their potential impact on information access and evidence synthesis is rapidly expanding.
In this 90-minute workshop, participants will first develop a foundational understanding of key terminology, core concepts, and the current landscape of agentic AI. We will then explore real-world examples demonstrating how agentic AI systems are being used in systematic reviews—including eligibility criteria development, search strategy, citation screening, data extraction, and validation workflows. We will focus on the application of search strategy creation and showcase how the audience can build a search strategy with the assistance of AI agents.
Finally, we will examine practical frameworks for critically appraising these systems, focusing on validity, performance claims, biases, and evaluation methodologies.
Participants will leave with:
- A clearer understanding of what AI agents and agentic systems are (and what they are not)
- Realistic expectations about their potential value and limitations in health libraries and evidence synthesis
- Practical skills to evaluate claims, ask informed questions, and support decision-making in their institutions
This session is designed for health librarians, information specialists, and anyone interested in the future of AI-supported evidence synthesis.
Background: GenAI-assisted search strategies are gaining momentum. Loon Lens is a Canadian initiative and the first-ever scientifically validated agentic AI platform for systematic literature reviews. Loon Lens supports key steps in evidence synthesis, including developing eligibility criteria, building search strategies, running literature searches within OpenAlex, and conducting title, abstract, and full-text screening, as well as data extraction. It is HTA-compliant, citable, and demonstrates high performance.
Methods: We tested Loon Lens as an independent comparator to generate multiple literature search strategies based on predefined inclusion criteria, and we compared its performance with that of an experienced information specialist and researcher. With humans kept in the loop, agentic AI is not intended to replace human expertise but has been shown to enhance efficiency in literature searching.
Results and discussion: We will share our experience using Loon Lens to refine literature search strategies and reduce human time investment in the systematic review process. We will also discuss the strengths and limitations of AI tools in literature searching. As a research assistant and valuable collaborator, Loon Lens—when used alongside librarians—has the potential to substantially strengthen the quality, consistency, and timeliness of future systematic review workflows.