An Eye Tracking Study Assessing How Developers Read Stack Overflow to Summarize APIs

Stack Overflow is a question-and-answer forum for software developers. This resource provides essential information to developers working on a variety of tasks, helping improve productivity. Despite its popularity, there is limited understanding of how developers read, understand, and extract information from Stack Overflow pages. This thesis presents an eye-tracking study designed to investigate the behaviors of software developers as they summarize API elements using information from Stack Overflow. Through a controlled experimental design involving 30 eye-tracking sessions, the research examines the specific content elements developers prioritize and how closely their summaries reflect the information they read. Results showed that developers predominantly focused on question texts and question code snippets. Semantic similarity analysis conducted using Sentence-BERT revealed a moderate alignment between the developers' summaries and the content they read on Stack Overflow. These findings imply that Stack Overflow's specific content structure can significantly support developer understanding and productivity. This highlights the potential for enhancing informal documentation and the design of future educational tools and resources for increasing developer efficiency.