Doctoral defence: Heidi Taveter "Using Programming-Process Data of Introductory Programming Courses: Finding Solver Types, Giving Feedback, and Plagiarism Detection"

On January 20 at 11:00 Heidi Taveter will defend her doctoral thesis "Using programming-process data of introductory programming courses: finding solver types, giving feedback, and detecting plagiarism" to obtain the degree of Doctor of Philosophy (in Computer Science).

Supervisor:
Assoc. Prof. Marina Lepp, University of Tartu

Opponents:
Prof. Erik Barendsen, Radboud University (The Netherlands)
Assoc. Prof. Julien Broisin, University of Toulouse (France)

Summary
Over time, programming has become a crucial skill learned by students from diverse backgrounds and disciplines with varying levels of prior experience, while high dropout rates remain a significant challenge. Therefore, it is essential to explore how students differ in learning approaches while programming and identify effective ways to support them, particularly at the beginning of the course. This thesis aimed to explore how programming-process data can be utilized to profile and support students in introductory programming courses and to identify plagiarism instances.

This research showed that students can be grouped into solver types based on their behavior patterns in programming, with similar patterns appearing among both beginners and non- beginners. A notable discovery is that the late start of the first execution of the program is associated with lower performance. Another finding is that some groups achieved comparable results despite differences in behavior patterns, which emphasizes the diversity. It is essential to note that, compared to the first midterm exam, groups had more notable differences in the second midterm scores, and solver types are not persistent during an introductory programming course. Furthermore, the feedback derived from programming log data decreased the time required by beginners for solving programming tasks and improved their exam test results, which focused on code reading skills. The study showed that, unlike programming style features, general style features are stable and well-suited for plagiarism detection based on log data.

Based on the results of this thesis, an important recommendation is to promote a teaching approach that encourages students to run their programs frequently as an integral part of the coding process. In addition, it is effective to use log-based feedback to improve beginners' code-reading abilities. It is essential in the era of AI, when it is increasingly necessary to develop strong code-reading skills.