SBIR Phase I: Identifying and Countering Misinformation on Closed Messaging Platforms (COVID-19)

小型企业创新研究(SBIR)第一阶段项目的更广泛影响将是推进在线错误信息检测、优先排序和响应的最新技术。随着越来越多的平台转向端到端加密，开发新的工具和算法以应对这些平台上的错误信息至关重要。WhatsApp、Viber、Line、Telegram和Signal等平台上的加密保护了数百万美国人的通信，但也可能允许谣言、错误信息、虚假信息和其他威胁传播。这个项目建立了“错误信息专线”：允许用户与领先的事实核查组织核实潜在错误信息的在线账户。拟议的算法将允许美国人与事实核查组织核实潜在的错误信息，使他们能够在网上识别危险的--有时是危及生命的--健康错误信息。该项目增强了在线通信平台的用户能力，提高了他们的安全性，同时保持了端到端加密的好处。这个小型企业创新研究(SBIR)第一阶段项目将克服关键挑战，使Tipline能够高效地扩展到数百万用户。Tipline需要人工智能和人类知识的深思熟虑的结合。同样的误导性说法经常在网上以许多不同的方式重复和重申。为了使Tiplines扩大规模，有必要识别提交给Tiplines的消息中提出的主张，并将它们与现有的大量事实核查进行匹配。索赔可以是文本、图像、视频或音频消息的形式，并以许多不同的人类语言表达。在这个项目中，该公司正在构建所需的算法，以跨语言和跨格式匹配这些声明。这一结果在自然语言处理、计算机视觉和机器学习方面开辟了新的天地。这一奖项反映了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 broader impact of this Small Business Innovation Research (SBIR) Phase I project will be to advance the state-of-the-art for detecting, prioritizing, and responding to misinformation online. As more platforms are moving to end-to-end encryption, it is critical that new tools and algorithms are developed to respond to misinformation on these platforms. Encryption on platforms such as WhatsApp, Viber, LINE, Telegram, and Signal protects the communications of millions of Americans but also potentially allows rumors, misinformation, disinformation, and other threats to spread. This project builds ‘misinformation tiplines’: online accounts that allow users to check potential misinformation with leading fact-checking organizations. The proposed algorithms will allow Americans to check potential misinformation with fact-checking organizations allowing them to identify dangerous—and for health misinformation sometimes life-threatening—misinformation online. The project empowers users of online communication platforms and improves their safety while maintaining the benefits of end-to-end encryption.This Small Business Innovation Research (SBIR) Phase I project will overcome key challenges in order to allow tiplines to scale efficiently to millions of users. Tiplines require a thoughtful combination of artificial intelligence and human knowledge. The same misleading claims are often repeated and restated in many different ways online. In order for tiplines to scale, it is necessary to identify the claims being made in messages submitted to tiplines and match them against the extensive library of existing fact-checks. Claims can be in the form of text, image, video, or audio messages and expressed in many different human languages. In this project the company is building the algorithms needed to match these claims across languages and across formats. The results break new ground in natural language processing, computer vision, and machine learning.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.