The impact of training-by-examples on inspection performance using two laboratory experiments

Abstract

Software inspection is often seen as a technique to produce quality software. It has been claimed that expertise is a key determinant in inspection performance particularly in individual detection and group meetings [38]. Uncertainty among reviewers during group meetings due to lack of expertise is seen as a weakness in inspection performance. One aspect of achieving expertise is through education or formal training. Recent theoretical frameworks in software inspection also support the idea of possible effects of training on inspection performance [38]. To investigate this further, two laboratory experiments were conducted to test the effects of training using defect examples. Our findings show conflicting results between the two experiments, indirectly highlighting the importance of an effective inspection process. The results have implications for the use of a repository of defect examples for training reviewers.

Investigating training effects on software reviews: a controlled experiment

“…Process training with worked examples, was used in [4,5]. They were motivated by the training literature from the education area.…”

“…Worked examples significantly increase the students' understanding about the knowledge taught during class [6,31]. Worked examples were examples of different types of defect introduced during the training session [4,5]. The contents of process training with worked examples include: • All of the contents of process training.…”

“…Other than [3,4,5], we are not aware that the effectiveness of inspection training has been empirically studied.…”

“…This study is chiefly motivated from previous experiments conducted [4,5]. They tested the effects of training using worked examples on inspection performance.…”

“…Naturally, any activity aimed at improving the performance of inspectors would be deemed favourable to both practitioners as well as to researchers. This study is motivated by previous work by Chowdhury and Land [4,5], on the effects of inspector training on inspection performance. There have been few studies in this area.…”