CAREER: SHF: Debugging Mental Models

当软件不能像预期的那样运行时，致力于解决这个问题的软件开发人员就会参与调试活动。调试通常是具有挑战性的，因为开发人员必须生成关于故障原因的可能假设，通过对复杂代码的行为进行推理，费力地测试假设，并使用调试工具来生成和识别相关的运行时状态。对软件用户来说，这些挑战可能表现为具有未修复缺陷的软件和增加的软件成本。此外，专家开发人员拥有的知识，如果提取出来，将有助于更容易地进行调试。长期以来，专家完成简单的编程任务的速度比非专家快十倍以上，这主要是因为他们以前见过类似的情况。该项目解决了调试的心理模型的本质问题，并为理解、外部化和共享对程序运行时行为的专家见解创建了科学基础。这个项目包含了一组全面的活动，用于指导软件开发人员的研究和设计新形式的调试工具。广泛的开发人员样本将被观察到，以收集从调试任务中捕获心理模型的草图，抽象公共构建块，并计算这些构建块的频率。新的调试工具类型将被设计出来，使开发人员能够通过记录、抽象和重放对程序执行的查询以及可视化程序执行中元素之间的关系来具体化他们的心智模型。新的调试工具将通过用户研究进行评估，以了解开发人员是否以及如何外部化和共享他们的调试心理模型。总之，这些活动将加深对开发人员如何调试和为开发人员更有效地工作创建新技术的科学理解。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

When software fails to behave as intended, a software developer working to address this issue engages in the activity of debugging. Debugging is often challenging, as developers must generate possible hypotheses about a failure's cause, laboriously test hypotheses by reasoning about the behavior of complex code, and use debugging tools to generate and identify relevant runtime state. These challenges may be manifest to software users as software with unfixed defects and increased software cost. In addition, expert developers have knowledge which, if extracted, would be valuable in helping to debug more easily. Experts have long been shown to do even simple programming tasks over ten times more quickly than non-experts, largely because they have seen similar situations before. This project addresses the question of the nature of mental models of debugging and the creation of a scientific foundation for understanding, externalizing, and sharing expert insight into the runtime behavior of programs.This project encompasses a comprehensive set of activities for conducting studies of software developers and designing new forms of debugging tools. A broad sample of developers will be observed to collect sketches capturing mental models from debugging tasks, abstract common building blocks, and count the frequency of these building blocks. New types of debugging tools will be designed which enable developers to externalize their mental models by recording, abstracting, and replaying queries over program executions and visualizing relationships between elements within program executions. New debugging tools will be evaluated through user studies to understand if and how developers may externalize and share their debugging mental models. Together, these activities will deepen the scientific understanding of how developers debug and create new techniques for developers to work more effectively.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.