Content area
The rapid evolution of computer science education calls for innovative strategies to support both students and teachers to ensure equitable learning opportunities and effective teaching methodologies. This dissertation contributes to the broader goal of improving computer science education by proposing novel data-driven approaches to identifying struggling students while strengthening teacher preparedness.
The first contribution of this dissertation focuses on leveraging the granularity of keystroke data to analyze students’ programming behaviors and identify early indicators of struggle. Programming assignments play a crucial role in developing problem-solving skills and computational thinking; however, students — especially in introductory level programming courses — frequently struggle with grasping the syntax and logical structure of their code leading to high dropout and failure rates. To address this challenge, this work introduces, From Typing to Insights, a code visualization tool that reconstructs students’ coding processes from keystroke logs and automatically generates execution logs against unit tests at different time intervals, offering deep insights into students’ coding habits, debugging patterns and logical progression in problem-solving.
Building upon this, the second contribution integrates keystroke analytics with self-reported student experiences to propose TrackIt, a novel rule-based detection system that analyzes students’ keystroke data to classify students into different struggle categories. TrackIt features a copy-paste detection functionality, which flags students who paste large portions of code, including those potentially from AI-generated tools. The tool enables a more precise understanding of students’ learning difficulties. Additionally, TrackIt, combined with thebaseline reports, helps identify the most challenging concepts in introductory programming courses, thereby providing instructors with valuable insights to refine instructional approaches.
The third contribution shifts focus to teacher development, addressing disparities in computer science education, particularly in rural and underserved communities. Although CS education has expanded in recent years, access remains uneven due in large part to a shortage of qualified instructors. We investigate how participation in a structured teacher training program influences teachers’ professional computing identities, commitment and overall confidence and competence in teaching computer science. Following training, the findings show that rural teachers reported positive changes in their identities and teaching competencies and are more likely to advocate for more students to take computer science courses. Teachers in rural areas also showed a marked improvement in confidence and commitment to teaching computer science.
Finally, to further understand teachers’ learning perspectives, the fourth contribution of this dissertation conducts in-depth study of teacher reflective journals, utilizing both human annotations and Large Language Model (LLM)-based analysis. By examining teachers’ insights into their learning experiences, instructional challenges and educational growth, this research contributes to the broader discourse on teacher development. These contributions present a comprehensive framework for advancing computer science education.