Rain D Librarian

Article: How Well Can ChatGPT Answer Library Reference Questions? – Choice 360

Authors: Sandy Hervieux and Amanda Wheatley are academic librarians with expertise in information literacy, artificial intelligence, and user services. Sandy is the Head of the Nahum Gelber Law Library at McGill University, and Amanda is an Associate Librarian at the same institution. Their research interests revolve around reference services, user experience, and the impact of AI on library services. 

#ChatGPT #LibraryReference #InformationLiteracy #AIinLibraries #Librarians #AcademicLibrarians #ResearchInterest

Summary

The rise of ChatGPT has sparked discussions among librarians regarding its potential impact on information validity, user evaluation, and the future of reference services. Two academic librarians decided to put ChatGPT to the test by asking it four reference questions based on their own virtual reference service. The goal was to assess its ability to provide accurate and helpful information.

Reference Question 1 – Dataset on Church Attendance Rates in the US (1950-2000): ChatGPT offered some valuable starting points for finding the dataset but lacked clarity on the curated datasets and where to find them. Additionally, it suggested sources like the National Center for Health Statistics, which needed to be more relevant. While it encouraged users to evaluate credibility, consulting a librarian for further assistance should have been recommended. ChatGPT provided some direction, but undergraduate students may require more guidance on this complex topic. 

Reference Question 2 – Obituary of Ruth Bader Ginsburg: ChatGPT demonstrated its limitations by lacking real-time internet access and providing only suggestions for finding the obituary. A librarian would be better suited to give an exact answer, but ChatGPT’s suggestions were reasonable. 

Reference Question 3 – Scientific Articles and Consumer Reports in the Travel Industry: ChatGPT recommended various databases but overlooked sources specifically for industry reports. It also mentioned TripAdvisor, which does not contain industry information. The guidance needed more details on searching for peer-reviewed journals and evaluating sources. 

Reference Question 4 – Primary Source on the Fall of the Berlin Wall: ChatGPT provided a valuable recommendation for a fictional primary source, showcasing its predictive nature. When asked about the book’s authors, it clarified its limitations and provided alternative primary sources. This interaction highlights the importance of teaching users to use ChatGPT effectively and continue critically evaluating sources. 

Conclusion: ChatGPT offers some assistance in answering reference questions, but its limitations, such as no real-time internet access and predictive responses, call for cautious usage. Librarians remain crucial in guiding users, offering specialized research skills, and ensuring information credibility. As librarians navigate the integration of AI, collaboration with AI tools can enhance the user experience. Still, human expertise remains essential for reliable and accurate information access.