Applying Generative AI to Analyze Statements from Student Surveys

Applications based on Generative Artificial Intelligence (GenAI) already permeate many areas of higher education. In this paper, we discuss using GenAI to assist staff in extracting information from qualitative data gathered by student surveys.Manually analysing surveys of large groups of students usually is a tedious task. In particular, categorising and evaluating open-ended questions can be very time consuming. As a means for effectively processing and analysing large amounts of student feedback, we experimented with using GenAI for this task. In this work, we describe a simple, inexpensive and easily adaptable setup for analysing student statements, and report on our experience with this approach. The goals are (1) to let GenAI assign each student statement to an adequate content category and to accurately determine its sentiment, and (2) to use GenAI to produce summaries of statements.We implemented our approach twice, using two different GenAI products, namely the commercial OpenAI API and a self-hosted Llama model. Comparing the results, we find that both are valuable in the process of evaluating students’ statements with respect to their sentiments and towards categorizing and summarizing them.