Yes, there will be additional presenters for this sessionCommittee: Group V: Access and Equity

First Learning Outcome: Understand the machine learning analysis approach to categorize and summarize written student comments 

Second Learning Outcome: Consider student perspectives on inclusive teaching practices and reflect on their own experiences

Third Learning Outcome: Brainstorm other data sources where this analysis approach would be beneficial

Core Competency: Interpretation and Application of Institutional and External Data

Proficiency: Records & Academic Services Proficiencies

Intended Audience: General Audience / Intended for Everyone

Collecting student feedback via open-ended survey responses provides rich data that can inform university policies and practices. However, analyzing and interpreting these large data sets can be time-consuming and is often a barrier to using student feedback. The University of Oregon (UO) is tackling this problem in order to make better use of the rich qualitative data we are already collecting. In this workshop, we will present how we have analyzed student comments about inclusive teaching practices that come from our recently re-designed course feedback surveys.

UO’s new Student Experience Survey asks students open-ended questions about 13 individual teaching practices upon completing a course. Each student selects the single teaching practice they consider most beneficial to their learning and the teaching practice that needs improvement in order to support their learning. Here, we focus on student comments about inclusive and accessible teaching practices to learn how students define these practices and which elements of their courses are the most beneficial or most in need of improvement to support their learning.

To identify themes from student responses, we used a hybrid approach combining traditional close reading with a machine learning approach called Latent Dirichlet Allocation (LDA), a topic modeling technique that identifies underlying themes in the text by looking for latent groupings of features. We used LDA as an exploratory tool to present a researcher with topics of student comments, which they analyzed and synthesized into the five themes presented here. The value of using LDA in this process comes from the neutrality of the algorithm’s topic groupings, which offers the researcher fresh ways of looking at the data that are relatively unconditioned by their prior expectation about what themes are present in student responses.

Presenter(s): 

Sung-Woo Cho University of Oregon

Austin Hocker University of Oregon

Grant Crider-Phillips University of Oregon