Gaze Collaboration Patterns of Successful and Unsuccessful Programming Pairs

Abstract

In this paper, we characterize the gaze collaboration patterns of successful and unsuccessful programming pairs as they traced and debugged fragments of code. A dual eye tracking experiment was performed on pairs of novice programmers and their fixation sequences were analyzed using Cross-Recurrence Quantification Analysis (CRQA), which is an analysis on cross-recurrence plots (CRP). Other eye tracking metrics were also used. Findings revealed that successful and unsuccessful pairs can be characterized distinctively based on their CRQA results, CRPs, and other eye tracking metrics. Successful pairs have more incidences of low CRQA reflected on their CRPs as single and isolated points, presence of more white bands and empty regions, and few rectangular segments known as laminar or “trapped” states. On the other hand, the unsuccessful pairs have more occurrences of high CRQA manifested on their CRPs as heavily clustered points and visually recurring patterns because of their more pronounced similar scan and fixation cluster patterns compared to successful pairs. Other eye tracking metrics also provided differentiation between the successful and unsuccessful pairs. These preliminary findings provide the groundwork to objectively quantify and characterize collaboration among programming pairs and can also be used in similar studies to strengthen pair programming.