Do friends collaborate and perform better?: A pair program tracing and debugging eye-tracking experiment

Document Type

Conference Proceeding

Publication Date

2018

Abstract

We characterize the extent of collaboration of pairs of novice programmers as they trace and debug fragments of code using cross-recurrence quantification analysis (CRQA). Specifically, we compare the collaboration of pairs whose composition is based solely on degree of acquaintanceship and look how

friendship affects collaboration and performance. Cross- recurrent plots (CRPs) were built using the pairs禽 eye tracking

data and CRQA metrics such as recurrence rate (RR), determinism (DET), average diagonal length (L), longest diagonal length (LMAX), entropy (ENTR), and laminarity (LAM) were derived from the CRPs using the CRP toolbox for MATLAB. The pairs禽 degree of collaboration was assessed based on these metrics. Findings reveal that the highly acquainted (HA) pairs collaborated better than the poorly acquainted (PA) pairs based on their percentage of recurrent fixations and matching scanpaths. However, the former禽s average performance score was lower than the latter. It was also observed that the PA pairs had steadier scanpaths compared to the HA pairs, which suggests that the PA pairs employed more logical and consistent debugging strategies that produced better scores than the HA pairs. Finally, the HA pairs were found to have struggled more in program comprehension as they had spent more time on certain regions of codes resulting to a lower average performance score compared to the PA pairs.

Share

COinS