Collaborative Research: DASS: Assessing Accountability of Tax Preparation Software Systems

随着美国税法的频繁变化，决策支持软件在帮助纳税人、专业人士和美国国税局（IRS）应对其复杂性方面发挥着至关重要的作用。税务筹划软件的使用大幅增加，2020年使用税务筹划软件的人数超过7200万人。然而，对税务软件的问责性进行评估的独立研究有限。本项目主要面临两大挑战：1)根据税法和专家的观点，确保软件的合规性、准确性和公平性；2)提高税务软件的测试、调试和补丁的可扩展性和准确性。该项目特别侧重于低收入纳税人的免税、抵免和减免，目的是确保所有纳税人，包括弱势社区的纳税人，缴纳税法规定的全部且仅缴纳税款。虽然人们期望具有法律和社会含义的软件应该是公平和符合法律的，但在法律关键领域（如税务软件）缺乏关于预期行为的正式规范，这对确保问责制构成了重大挑战。由于美国税法遵循先例的法律原则（先从先例），本项目提出，这些规范自然地以在给定环境中被认为相似的个人之间的变质关系存在。项目团队计划1)从美国税法中大量具有挑战性的需求中解释变形关系，2)通过利用与相关性相关的感知心理学原理自动提取这种变形关系，3)开发用于评估税务软件的工件，利用形式化验证技术利用这些关系。4)实验比较了税务软件与人类税务专家的准确性和对程序公正的感知。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。

As the U.S. tax law frequently changes, decision-support software plays a crucial role in helping taxpayers, professionals, and the Internal Revenue Service (IRS) navigate its complexities. The use of tax preparation software has witnessed a significant increase, with over 72 million people utilizing it in 2020. However, there has been limited independent research conducted to assess the accountability of tax software. This project takes on two main challenges: i) ensuring the software's compliance, accuracy, and fairness based on tax law and experts' perspectives, and ii) enhancing scalability and precision in testing, debugging, and patching tax software. The project specifically focuses on exemptions, credits, and deductions for low-income taxpayers, with the aim of ensuring that all taxpayers, including those from vulnerable communities, pay all and only the taxes that tax law prescribes.While it is expected that software with legal and social implications should be fair and compliant with the law, the absence of formal specifications regarding expected behaviors in legal-critical domains like tax software poses significant challenges in ensuring accountability. Since U.S. tax law adheres to the legal doctrine of precedent (stare decisis), this project proposes that these specifications naturally exist as metamorphic relationships between individuals who are considered similar within a given context. The project team plans to 1) explicate metamorphic relations from a large set of challenging requirements in U.S. tax law, 2) automate the extraction of such metamorphic relations by leveraging principles from the psychology of perception pertaining to relateness, 3) develop artifacts for assessing tax software that leverages those relations using formal verification techniques, and 4) experimentally compare the tax software's accuracy and perception of procedural justice to that of human tax experts.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.