Held in Pittsburgh, Pennsylvania from April 30 – May 2, the annual Medical Library Association conference was great as always, despite having to make some last mintue pivots due to travel restrictions for employees of the NIH. One major pivot was having to find a replacement for our chosen main speaker. The planning committee asked a group of medical students from StreetMedicine@Pitt to fill in, and they shared their experiences about the “student-run interdisciplinary organization that strives to bring healthcare and social support to the rough-sleeping and unhoused community in Pittsburgh.” Hearing the students describe the Street Medicine team was very inspiring as they shared stories about the rounds they make every Wednesday and the connections they make with the community.
From the sessions I chose to attend, three major themes emerged: systematic review services and support; AI impact and ethical concerns; and inclusive hiring practices and onboarding. One interesting place two of the themes overlapped was in a paper session called “Anything You Can Do, AI Can Do Better… or Can It? Comparing ChatGPT’s Search Strategy Outputs with Cochrane Review Searches” presented by a group of librarians from UNC Chapel Hill. The librarians studied ChatGPT’s current capabilities to produce comprehensive searches for systematic reviews. They compared human created searches to GenAI produced searches, using nine published Cochrane reviews for the comparisons. They did find a few pros but many cons to using ChatGPT to produce comprehensive searches. The pros: it was a bit of a time saver, and there were very few errors with syntax and logic. Cons included: fake MeSH terms, duplicative and lengthy keyword search phrases, lower recall than human searches, inability to reliably access and replicate validated filters, and high variability of results based on models and prompts. Their research confirms what I have seen using GenAI in research – it can be helpful to give broad outlines or to get started on a project, but human intervention is still necessary. The full slide presentation is available online.
Many sessions surrounding AI were focused on (or at least touched on) ethical issues and environmental impact of Generative AI. It was clear that many librarians are very concerned about the use of AI. One librarian, calling herself A Librarian Against AI, shared a zine she created that covers many of these issues. In a separate presentation, one librarian gave tips to help mitigate the environmental impact of AI use, such as: utilizing AI tools to fit the task, like pre-trained small models for simpler tasks; using efficient prompting; and avoiding generating unnecessary images. This one especially hit me – I learned that image generation uses three times the energy of a GenAI text inference, which already requires a lot of energy.
I truly enjoy attending the annual MLA conference, this was the third year I’ve gone. I learn so much as a health sciences librarian and have fun networking. Finding people who also nerd out about systematic reviews is always a great time. And, I like to treat myself to local food. If you are in Pittsburgh in the near future, I highly recommend Bar Marco. MLA 2026 will be in Milwaukee, and I’ll certainly be attending!
Submitted by Lynn Warner, University of Cincinnati