Convergence Accelerator Phase I (RAISE): Credible Open Knowledge Network

NSF融合加速器支持基于团队的多学科努力，以应对国家重要性的挑战，并在不久的将来显示出交付成果的潜力。此融合加速器第一阶段项目的更广泛影响和潜在的社会益处是开发新的能力，以确保从称为知识图谱或知识网络的大型数据集合中收集的信息的质量和可信度。今天的许多智能软件产品都是由大量的知识集合提供支持的，而这些知识集合往往是专有的。基于公共数据开发开放可用的基础设施是2019年整体融合加速器试点的主要路线之一。这一具体项目的目标是确保在发展这样一个开放的知识网络方面的可信度--完整性、完整性和真实性。因此，该项目的努力可能会为2019年启动的许多融合加速器第一阶段工作提供见解和价值。该项目计划开发一个揭露错误信息的资源，这些错误信息对个人、组织、社区和政策制定者的决策非常重要。该项目的初始使用案例将是医疗保健，帮助创建方法，确保收集的与健康相关的数据是可靠的，以支持医疗保健理解和决策，并通过查看网络威胁情报集合中的危害指标的有效性和可靠性来缓解来自软件漏洞的安全威胁。该项目支持一个多机构和多学科的团队，该团队拥有横跨计算机科学、经济学、新闻和传播、政治学、心理学和公共卫生的宝贵专业知识。该团队包括来自行业合作伙伴、政府研究组织、学术机构和具有与信息可信度相关的互补专业知识的国际组织的研究人员。该项目的第一阶段将侧重于团队组建、研究规划和开发概念验证。如果项目成功进入第二阶段，它将建立一个由数据集、算法、软件以及利益相关者社区组成的可持续生态系统，以便在开放的知识网络中创建和充分利用可信度工具。该团队将对几个交叉问题进行受使用启发的趋同研究：(1)建立可信知识图谱的模型，决定为了提高可信度需要获取哪些类型的知识，以及这些知识应该如何表示，以便能够采用计算方法。(2)发展数据驱动的理解，了解哪些因素有助于事实陈述的说服力，哪些信号有助于衡量可信度，以及如何利用这些信息来设想针对错误信息的对策。(3)设计计算方法，利用知识图来审查语句并生成可验证和可靠的解释。(4)随着时间的推移，启动、发展和维持可信的开放知识网络的程序和机制。这项研究的发现有可能显著影响对信息和错误信息的创建和消费的理解，并可能引发新的调查路线，同时还有助于在知识网络中总体上创建更可靠的信息。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

The NSF Convergence Accelerator supports team-based, multidisciplinary efforts that address challenges of national importance and show potential for deliverables in the near future.The broader impact and potential societal benefit of this Convergence Accelerator Phase I project is to develop new capabilities to ensure the quality and credibility of information collected in large assemblages of data known as knowledge graphs or knowledge networks. Many of today's intelligent software products are powered by massive knowledge assemblages which are often proprietary. Developing an openly available infrastructure based on public data is one of the overarching tracks of the overall Convergence Accelerator Pilot in 2019. The goal of this specific project is to ensure credibility -- integrity, completeness and truthfulness -- in developing such an open knowledge network. As such, this project's efforts are likely to provide insights and value to many of the Convergence Accelerator Phase I efforts initiated in 2019. This project plans to develop a resource for debunking misinformation, which is important to decision making by individuals, organizations, communities, and policy makers. The project's initial use-cases will be healthcare, helping to create methodologies that ensure that health-related data assembled is reliable to support healthcare understanding and decision making, and mitigating security threats from software vulnerabilities by looking at the validity and reliability of indicators of compromise within collections of cyber threat intelligence. The project supports a multi-institutional and multidisciplinary team that has valuable expertise spanning computer science, economics, journalism and communication, political science, psychology, and public health. The team includes researchers from industrial partners, government research organizations, academic institutions, and international organizations with complementary expertise relevant to information credibility. Phase I of the project will focus on team formation, research planning, and developing proof-of-concept. If the project successfully proceeds to phase II it would develop a sustainable ecosystem of datasets, algorithms, software, as well as a stakeholder community for creating and fully utilizing credibility tools across an open knowledge network.The team will conduct use-inspired, convergence research on several cross-cutting problems: (1) Modeling credible knowledge graphs, deciding what types of knowledge need to be captured in order to promote credibility, and how such knowledge should be represented in order to enable computational approaches. (2) Developing data-driven understanding of what factors contribute to the persuasiveness of factual statements, what signals help gauge credibility, and how to employ that information to envision countermeasures against misinformation. (3) Designing computational methods that leverage knowledge graphs to vet statements and generate verifiable and reliable explanations. (4) Procedures and mechanisms for initiating, developing, and maintaining a credible open knowledge network over time. The findings of this research have the potential to significantly impact understanding of information and mis-information creation and consumption and may trigger new lines of investigation while also helping to create more reliable information in knowledge networks generally.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.