CAREER: Empirical Investigations of Software Inspections

该职业奖支持软件检验方法的实证评估的研究和指导。软件检查是一种广泛使用但非常昂贵的验证过程。最初的研究表明，这一过程背后的许多假设是不正确的，因此，本研究涉及以下三个实验家族。(1)缺陷检测技术比较。本实验比较了系统检查技术与标准Ad Hoc和Checklist方法的成本效益。它的假设是系统检查技术比目前使用的非系统检查技术更有效。(2)成本效益研究。本实验是在一个实际的开发项目中进行的，以比较几种不同的检查方法在最小检查间隔和最大有效性之间的权衡。它的假设是，最近提出的几种方法显著增加了间隔，但只提供了轻微的改善有效性。(3)检验进入标准。这个实验试图确定工件的质量是否会影响检测的有效性。它的假设是传统的检查对于缺陷密度非常高或非常低的工件是无效的。这项研究也为研究生和本科生的教学提供了材料。这包括一个为期4周的教学模块，学生在其中进行实验，收集和分析数据，并测试假设作为研究生软件工程课程的一部分。目前正在开设一门本科软件开发课程，让所有学生一起为有实际需求的实际客户开发一个大规模应用程序。

This CAREER award supports research and instruction in the empirical evaluation of software inspection methods. Software inspection is a widely used, but very expensive validation process. Initial research has demonstrated that many of the assumptions underlying this process are incorrect and, therefore, this research involves the following three families of experiments. (1) Comparing defect detection techniques. This experiment compares the cost-effectiveness of systematic inspection techniques with that of the standard Ad Hoc and Checklist approaches. Its hypothesis is that systematic inspection techniques are more effective that the nonsystematic ones currently being used. (2) Cost-benefit studies. This experiment is conducted on a live development project to compare the tradeoffs between minimum inspection interval and maximum effectiveness for several different inspection approaches. Its hypothesis is that several recently proposed methods significantly increase interval, but provide only a slight improvement in effectiveness. (3) Inspection entry criteria. This experiment attempts to establish whether or not an artifact's quality affects inspection effectiveness. Its hypothesis is that traditional inspections are not effective for artifacts with very high or very low defect density. This research is also producing materials that enhance both graduate and undergraduate teaching. This includes a 4-week teaching module in which students perform experiments, collect and analyze data, and test hypotheses as part of a graduate software engineering course. An undergraduate software development course is being developed in which all students work together to develop one large-scale application, for real customers with real needs.