Impact of AI as a Tool in University Data Visualisation Courses

Abstract

Data visualisation comprehension is a critical skill in our society and acquisition of the needed skills remains challenging. How AI tools can be used to ease some of these challenges is answered by this thesis by exploring the perceptions of students of using AI tools and evaluating impact of using AI tools during university data visualisation courses on students’ data visualisation literacy. The study employs a multi-step approach. Evaluation of DVL of large language models (LLMs) was done in a pre-study by using the Critical Thinking Assessment for Literacy in Visualisations (CALVI) test. Capabilities of LLMs were found generally inferior to human students (with lower distance in misleading charts). The main research was split in two parts with students from undergraduate and graduate courses. The first part used a pre/post test design with CALVI to measure impact of students using AI on the DVL change of consumption and critical thinking aspects, while the second part evaluated the impact of AI on practical creation of digital data visualisations by programming exercises. In both parts, qualitative data to explore students’ perceptions and usage patterns was collected via surveys and interviews. Assumptions from the pre-study and the development of DVL of AI models were finally updated in a follow-up of the pre-study, which found that capabilities of LLMs are rapidly increasing with newest models being able to surpass human scores. The results were not generally conclusive. While the first part did not find a statistically significant difference between students who used AI and those who didn’t, the second part showed a clear advantage of AI usage in creating data visualisation in programming exercises. Students’ perceptions of AI as a useful tool in easing technical and programming related problems is contrasted by a lower number of students answering, that they felt they learnt a lot with the task. The study concludes that AI can be a useful tool in data visualisation creation tasks in university data visualisation courses, but adaption of exercises and lectures may be needed to fully employ resources freed by AI usage or it could lead to a degradation of student’s sense of learning from the tasks.

Date of Award	2025
Original language	English
Supervisor	Mandy Keck (Supervisor)