From Transcriptions to Memory: Libraries as Curators of Community Questions in the Age of AI

Abstract

Libraries have long preserved recorded knowledge, but not the questions patrons ask. Every reference interaction produces a question, a moment of curiosity or uncertainty expressed in natural language. Current practice reduces these questions to transaction counts and category labels. Recent studies have applied machine learning and natural language processing to reference transcripts, mainly to classify question types or improve service workflows, not to preserve them as culturally meaningful data.

Drawing on a reconceptualization of librarianship as the facilitation of knowledge creation through conversation, we propose that libraries are equally positioned to preserve traces of that conversation. Patron questions, collected and analyzed over time, may constitute a form of community memory, revealing shifts in community concern and gaps between institutional knowledge and public understanding. During a pandemic, for example, patron questions expose uncertainties that published documents rarely capture.

We outline a conceptual framework for question-based community memory, discuss how AI can support structuring and interpreting such data at scale, and address ethical considerations around patron privacy.