Get Full Government Meeting Transcripts, Videos, & Alerts Forever!
Get email alerts on the Ai Searching topic
No spam. Unsubscribe anytime.
AI can supplement — but not replace — expert systematic literature searching, presenters say
Summary
Kyle Holland said LLMs can help identify seed studies, extract searchable concepts and expand database choices, but they cannot replace human expert searchers; he recommended human peer review and careful prompting to improve search term collection.
Kyle Holland, an information specialist with the Center for Systematic Reviews and Research Synthesis at the Texas A and M University Medical Sciences Library, told webinar attendees that, in his view, “AI cannot outsearch you as an information professional.” He said large language models (LLMs) and generative systems can be a useful supplemental source for seed studies, synonyms and framing concepts, but that any AI-generated searches require human scrutiny and revision.
Holland described practical uses where LLMs can help: locating a handful of potentially relevant seed studies, identifying searchable concepts from those seeds, and suggesting term variations (including pluralization and grammatical forms) that human searchers might otherwise miss. He warned that LLMs are conversational and may “offer their evaluation” of evidence availability; searchers should treat that feedback skeptically.
Why it matters: Systematic searches underpin the completeness of evidence reviews. Holland recommended using AI to inform or refine searches rather than to compose final queries. He also noted that AI-based search constructors that produce executable queries are emerging (often targeted to open-access APIs such as PubMed) and can be helpful for scoping but still need expert review.
Practical guidance: Provide seed studies to a focused/customizable LLM for concept extraction, use AI to expand term lists and test AI-generated queries against expert-crafted searches, and always subject AI outputs to peer review before deployment in a formal review.
The webinar closed with a Q&A and resources list; the presenters said slides and a recording will be posted on the NNLM channel.
