Improving Model Inspection with Crowdsourcing

Abstract

Traditional Software Inspection is a well-established approach to identify defects in software artifacts and models early and efficiently. However, insufficient method and tool support hinder efficient defect detection in large software models. Recent Human Computation and Crowdsourcing processes may help to overcome this limitation by splitting complex inspection artifacts into smaller parts including a better control over defect detection tasks and increasing the scalability of inspection tasks. Therefore, we introduce a Crowdsourcing-Based Inspection (CSI) process with tool support with focus on inspection teams and the quality of defect detection. We evaluate the CSI process in a feasibility study involving 63 inspectors using the CSI process and 12 inspectors using a traditional best-practice inspection process. The CSI process was found useful by the participants. Although the preliminary results of the study were promising, the CSI process should be further investigated with typical large software engineering models.