Honors Program
Honors Program: Embargoed Theses
First Advisor
Bonita Sharif
Date of this Version
Spring 5-13-2025
Document Type
Thesis
Citation
Schoneweis, E. 2025. How Developers Read Bugzilla Reports To Summarize APIs: An Eye Tracking Study. Undergraduate Honors Thesis. University of Nebraska-Lincoln.
Abstract
Developers rely on detailed bug reports from websites like Bugzilla to understand, reproduce, and remedy software defects, yet the cognitive processes underlying how they read and summarize these reports remain underexplored. This thesis presents an empirical eye-tracking study with 30 participants—including industry professionals and students—who read real Bugzilla bug reports and produced concise written summaries. We capture fixations on key report elements (questions, answers, attachments) and compute gaze metrics (fixation counts, dwell times), linking these to summary correctness and the proportion of content used from pages via SBERT-based cosine similarity. Results show that participants allocate over 50 of fixations to answer sections, and longer dwell times on answers significantly predict fully correct summaries (U=2140, p=0.001), whereas raw quantity of cited content does not distinguish accuracy. Clustering reveals two effective reading–summarization strategies—shallow–quick and deep–verbose—yielding comparable accuracy, and task-level analysis across eight bug-report classes uncovers heterogeneity in engagement and accuracy not explained by gaze or structural metrics. Our findings inform the design of gaze-aware summarization tools, bug-tracker interfaces, and developer training by emphasizing selective attention to solution-relevant content. This work bridges eye-tracking research and bug-report summarization, offering novel insights into developer comprehension and guidance for automated support systems.
Comments
Copyright Elijah Schoneweis 2025