Developing Eye Tracking Methods for Detecting Source Code Comprehension Strategies

Abstract

Knowledge in STEM disciplines is becoming increasingly crucial to be able to use, understand, and create technology in the emerging society. Understanding technology, its importance, and its disadvantages are essential for participation in an emerging society. Adapting the education of children should be a key goal to prepare them for a hardly predictable future. As one domain of STEM, learning and teaching programming are considered difficult and pose great obstacles for both teachers and learners. Introductory programming courses are largely not as effective and successful as they could be, and according to extensive research in computer science education, one result is that students are not learning to program at all. Therefore, learner support is a vital part of introductory courses to help them cope with the technical and semantic problems they face. Modified programming languages, instructional design, and intelligent tutoring systems (ITSs) are possible measures of positively changing the outcome of programming courses. Eye tracking as technology is suitable for analyzing the comprehension process of reading source code, and it is a technological foundation to interactively support learners. Many research projects focus on the differences between novices and experts regarding their eye movement strategies and patterns. To date, specialized methods for analyzing and visualizing eye movement patterns, e.g., intending to detect useful data for a support system, are missing. Therefore, to move the knowledge in the source code comprehension community forward, this thesis focuses on developing and testing specialized visualization and analytical methods and tools. The research approaches and ideas necessary for this step are used in three case studies to test new visualizations, analysis methods, models, and learning hints. The results of these case studies serve as contributions to the source code comprehension research community and the vision of a dynamic learner support system. The contributions of this thesis primarily aim at better detecting source code comprehension strategies, as well as identifying methods and tools to support this detection.