The problem of abundance: Text mining approaches to qualitative assessment of asynchronous library instruction

Abstract

The expansive adoption of asynchronous library instruction in recent years remains strong even after the conclusion of most COVID-19 emergency remote teaching. This continued growth has also introduced unique methodological challenges for assessment of learning and instruction in the form of vast, unwieldy evaluation data. This study introduces the use of text mining, topic modeling, and exploratory data analysis (EDA) as a novel approach to address the evolving needs of the field of large-scale online library instruction assessment. Librarians employed computational methods on a corpus of 21,506 words from student survey responses representing the instruction experiences of 3720 students to evaluate initial effectiveness of a new instruction approach. The application of a Bayesian topic model revealed latent patterns in self-reported perceptions of course design, mastery of concepts, and gaps in learning. In contrast with the inherent limitations of traditional qualitative analysis, our computational-grounded approach highlights otherwise indiscernible trends and key issues in user experience, instructional content, and online teaching methods. Computational analysis offers the scalability and sustainability necessary to assess asynchronous instruction, and provides clear themes to inform decision-making.