Collaborative Research: RI: III: SHF: Small: Multi-Stakeholder Decision Making: Qualitative Preference Languages, Interactive Reasoning, and Explanation

在产品设计、公共政策、医疗保健、信息安全和隐私等广泛的应用程序中，表达和推理偏好的能力对于理性决策至关重要。由于在许多实际环境中缺乏定量偏好，人们对用定性偏好表示和推理的方法越来越感兴趣。此外，实际的决策制定场景通常涉及多个干系人，可能有相互冲突的偏好，并且一些干系人的偏好有时可能超越其他干系人的偏好，例如，由于干系人在组织中的相对位置。然而，现有的首选项语言和方法仅限于单个涉众设置。在此背景下，本项目汇集了一组在形式化方法、人工智能和偏好推理方面具有互补专业知识的研究人员，以开发用于表示和推理多利益相关者偏好的方法和工具。该项目产生的实用的开源多利益相关者决策支持工具将大大降低人工智能和正式方法在多个领域中用于多利益相关者决策的应用障碍。该项目加强了对ISU和PSU的研究生和本科生（包括女性和其他代表性不足的群体成员）在人工智能、正式方法和国家重要相关领域的研究培训。广泛传播研究成果（包括出版物、开源软件、数据、教程、课程材料），将研究成果纳入计算机科学、信息科学与技术、数据科学及相关学科的本科和研究生课程，并扩展到卫生、公共政策、安全和隐私等目标应用领域；这将受益于多利益相关者决策的先进工具，进一步增强项目的更广泛影响。该项目的主要智力价值集中在语言、算法和多利益相关者表示和偏好推理的软件方面当前最先进的实质性进展。研究人员将发展广义条件相对重要性和偏好理论（GCRIPT），这是一种包含现有偏好语言的多利益相关者偏好表达语言。由此产生的偏好推理器将能够(a)分析GCRIPT中表达的偏好，(b)根据多个利益相关者的偏好进行推理，不仅考虑他们的个人偏好，还考虑优先考虑某些利益相关者偏好的层次结构，以及(c)提供易于理解的偏好选择解释及其对利益相关者的影响。该项目还将增强构成偏好推理框架核心技术的底层模型检查技术；例如，在增量模型检查、反例分析和论证方面。由此产生的知识表示和形式化方法的进步有助于人工智能系统大大增强和扩展人类在多利益相关者决策中的能力。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。

The ability to express and reason about preferences over a set of alternatives is central to rational decision-making in a broad range of applications, such as product design, public policy, health care, information security, and privacy, among others. Because of the lack of quantitative preferences in many practical settings, there is increasing interest in methods for representing and reasoning with qualitative preferences. Furthermore, practical decision making scenarios typically involve multiple stakeholders, with possibly conflicting preferences, and the preferences of some stakeholders may sometimes override those of others, e.g., because of the relative positions of the stakeholders within an organization. However, existing preference languages and methods are limited to the single stakeholder setting. Against this background, this project brings together a team of researchers with complementary expertise in formal methods, artificial intelligence, and preference reasoning to develop methods and tools for representing and reasoning with multi-stakeholder preferences. The practical open-source multi-stakeholder decision support tools resulting from the project will significantly lower the barrier to the applications of AI and formal methods for multi-stakeholder decision making in a number of domains. The project enhances research-based training of graduate and undergraduate students, including females and members of other under-represented groups, at ISU and PSU in artificial intelligence, formal methods, and related areas of national importance. Broad dissemination of research results (including publications, open source software, data, tutorials, course materials), incorporation of research results into undergraduate and graduate curricula in Computer Science, Information Sciences and Technology, Data Sciences, and related disciplines, and outreach to targeted application domains e.g., health, public policy, security and privacy, that would benefit from advanced tools for multi-stakeholder decision-making further enhance the broader impacts of the project.The primary intellectual merit of the project centers around substantial advances in the current state-of-the-art in languages, algorithms, and software for multi-stakeholder representation and reasoning with preferences. The researchers will develop Generalized Conditional Relative Importance and Preference Theory (GCRIPT), an expressive language for multi-stakeholder preference representation that subsumes existing preference languages. The resulting preference reasoners will be able to (a) analyze preferences expressed in GCRIPT, (b) reason with the preferences of multiple stakeholders, taking into account not only their individual preferences, but also hierarchies that give precedence to the preferences of some stakeholders over those of others, and (c) offer easy-to-understand explanations of the preferred choices as well as their impacts on the stakeholders. The project will also enhance the underlying model checking techniques that form the core technology for the preference reasoning framework; e.g., in the areas of incremental model checking, counter-example analysis and justification. The resulting advances in knowledge representation and formal methods contribute to AI systems that substantially augment and extend human capabilities in multi-stakeholder decision making.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.