An Experiment to Investigate Interacting versus Nominal Groups in Software Inspection

Author(s): Diane Kelly, Terry Shepard
Venue: Proceedings of the 2003 conference of the Centre for Advanced Studies on Collaborative research
Date: 2003

Type of Experiement: Other
Sample Size: 7
Class/Experience Level: Graduate Student
Participant Selection: coursework
Data Collection Method: Observation, Survey, Project Artifact(s)

Quality
3

Link: http://delivery.acm.org/10.1145/970000/961343/p122-kelly.pdf?key1=961343&key2=6499003321&coll=GUIDE&dl=GUIDE&CFID=17961922&CFTOKEN=43734613

This paper investigates the effectiveness of group inspections over individual inspections. The participants were split into 3 groups based on experience levels and two rounds of inspection were conducted. Each round consisted of the individual members inspecting the code and noting defects, followed by the members of the group meeting to further inspect the document.

Two metrics were considered, meeting loss and meeting gain. Meeting loss occurs when an individual finds a defect that is not found in the subsequent group meeting. Conversely, meeting gain occurs when the group discovers a defect not found by the individuals. The results indicate that 27% of the defects found by individuals became meeting losses (not rediscovered in group). Also, 17% of defects found during the group inspection were meeting gains.

The results suggest that individual inspections are more productive at finding defects than group inspections. While additional defects are discovered during group meetings, it may not be sufficient enough to warrant the extra time required. However, the authors point out that the low number of participants disallow any statistical analysis of the results. Thus, they state that additional experiments are needed to validate any claims made in this paper.

0