An empirical investigation on the impact of training-by-examples on inspection performance

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 [33]. Uncertainty, among reviewers during the 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 [33]. A laboratory experiment was conducted to test the effects of training by examples on requirements inspection. Our findings show that the trained group performs significantly better than the group which received no training. However, the 'experienced' reviewers did not outperform those with no experience. The results have implications the use of a repository of defect examples for training reviewers. © Springer-Verlag Berlin Heidelberg 2004.