Abstract
This study shows a process of designing and comparing two quantitative evaluation methods for the structure of assurance cases (AC), which use two description methods: Goal Structuring Notation, and Dependability Cases. The International Organization for Standardization has mandated organizations to include AC in ISO 26262 and recommended to include AC in other standards. This study proposes two quantitative evaluation methods designed by authors, and assesses the necessity and effectiveness of the methods based on the results of quantitative evaluation of AC to confirm if the methods are useful for work improvement. We used a questionnaire completed by third party employees who are 1) working in an organization with written work procedures developed in Japanese organizational culture, 2) working in the manufacturing industry, and 3) engaged in daily routine work. These conditions define how written work procedures and documents have been created at each participant's workplace. Since we use these written materials as evidence to evaluate ACs, they satisfy certain conditions for preparation of the written information. After showing previous research on assurance cases and presenting a procedure of quantitative evaluation, we propose two equations for quantitatively evaluating a sub-goal. One equation simply integrates the numbers of Evidence and Monitoring. The second equation takes into account users' opinion on Evidence and Monitoring to quantitatively evaluate an optional sub-goal. We then describe and discuss the results of a questionnaire on the two quantitative evaluation methods for assurance cases, one using Eq. 1 and the other using Eq. 2. Responses were given quantitatively on a seven-point ordinal scale and qualitatively in a free descriptive space. We then assessed effectiveness and necessity of the two methods. The results were statistically significant for both "effectiveness" and "necessity," And we found that Eq.1 suggested more efficacy than Eq.2. This paper concludes with future research topics.
