Human Error Scenario Analysis of Software Defects

The detection and removal of software defects have been causing substantial global costs, yet safety-critical software systems persistently encounter residual defects leading to occasional accidents. Among the diverse causes of defects, human errors of developers stand out significantly. Despite the existence of human error taxonomies in software engineering, the cognitive processes underlying how these errors contribute to defects remain largely unexplored. This paper proposes a novel methodology, Human Error Scenario Analysis (HESA), aimed at reconstructing the cognitive mechanisms behind software defects. Grounded in two fundamental psychological principles--the presence of underlying cognitive mechanisms in human errors and the influence of contextual factors--HESA offers a fresh perspective for understanding how software defects are caused by human errors. A case study was conducted to assess the feasibility and reliability of the proposed approach, involving 3 analysts and 120 defects across 6 industrial projects. Results indicate that using HESA, analysts were able to identify at least one human error mode behind 98% of the defects, and reached an interrater reliability of 0.61, which falls within the range of Substantial Agreement according to the commonly used interpretation criteria for interrater reliability. This underscores the potential of HESA in comprehending the pathways through which human errors engender software defects, thereby facilitating proactive defect prevention strategies.