Computer-mediated group support, anonymity, and the software inspection process: an empirical investigation

Abstract

In software inspection, a key principle endorsed by Fagan (1986) is openness. However, scholars have recently questioned the efficacy of openness. For example, some argue that ego-involvement and personality conflicts that become more transparent due to openness might impede inspection. Still others point out that familiarity and (preexisting) relationships among inspection team members negatively affect the comprehensiveness in detection of defects. This brings up concerns if the openness as originally envisioned by Fagan may in fact lead to suboptimal outcomes. As the trend towards computer-based inspection continues, we believe that anonymity could play a positive role in overcoming some of the drawbacks noted in team-based inspection. Drawing upon the literature on software inspection and group support systems, this research proposes possible influences of group member anonymity on the outcome of computer-mediated software inspection and empirically examines the validity of the posited relationships in a set of controlled laboratory experiments. Two different inspection tasks with varying levels of software code complexity are employed. While both the control groups (i.e., teams without anonymity) and treatment groups (i.e., teams with support for anonymity) consume more or less the same time in performing the inspection tasks, the treatment groups are more effective in identifying the seeded errors in the more complex task. Treatment groups also express a more positive attitude toward both code inspection tasks. The findings of the study suggest a number of directions for future research.