Characterizing Individual Gaze Patterns of Pair Programming Participants

Abstract

In this paper, we investigate the gaze patterns of individuals within pairs as they traced and debugged fragments of code. We performed a dual eye tracking experiment, recorded their fixations and computed the gaze-based metrics of these individuals. The participants within pairs were categorized into a more successful and less successful participant based on the number of bugs found. Results suggest that the more successful participants acquaint themselves first with the program encoding more information about the program, have more increased attention on the erroneous lines of code, strike a good balance between processing and searching, read code less linearly, and are more engaged in the task. The goal of this study is to capture individual expertise to gain insights on what makes the more skilled participants in a pair programming setup efficient and effective through their gaze patterns.