NRT-AI-FW-HTF: Co-Design of Trustworthy AI and Future Work Systems

NRT-AI-FW-HTF：值得信赖的人工智能和未来工作系统的协同设计

• 批准号: 2125677
• 负责人: Zoe Szajnfarber
• 金额: $ 300万
• 依托单位: George Washington University
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2021
• 资助国家: 美国
• 起止时间: 2021-09-01 至 2026-08-31
• 项目状态: 未结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2125677&HistoricalAwards=false
• 关键词: NRT; AI; FW; HTF; Co

The nature and structure of work are fundamentally changing as artificial intelligence (AI) becomes more deeply integrated within the structures of modern workplaces. This integration creates tension between the opportunities for ubiquitous AI to transform the workplace and emerging risks around bias, security, and privacy. Currently, AI tools are being developed at an unusually rapid pace, and deployed into environments where value maximization precedes regulation. The next generation of innovators accordingly needs a new kind of training. For algorithm designers, this means understanding and being sensitive to the context in which their creations may operate in unexpected ways through interaction with users in socio-technical ecosystems. For system designers, this means knowing enough about how AI tools are evolving to reimagine how tasks and processes could and should transform work in ways that fully leverage the potential power of AI tools. This National Science Foundation Research Traineeship (NRT) award to the George Washington University will address these needs by training doctoral students, master’s students, and graduate certificate students who will be prepared to make convergent research contributions to AI in the future workplace in a way that positively impacts society. The project anticipates training one-hundred and twenty (120) students, including twenty-five (25) funded Ph.D. trainees, primarily serving students in the discipline of computer science and systems engineering but with close interaction with the students and faculty in law, media, public affairs, public health, and international affairs.This NRT aims to educate researchers capable of “co-designing” AI algorithms and work systems to unlock new opportunities in both the capabilities of new systems and their “trustworthiness.” To accomplish this, the educational program aims to instill the following: 1) Comfort in bridging distant disciplines. Through novel onboarding sequences and shared experience of cross-disciplinary engagement with peers, mentors, and industry, the program will educate interdisciplinary, “comb-shaped” scholars who have a solid base in either AI algorithms or work system design and are also comfortable engaging deeply with other disciplinary areas fundamental to their chosen research problems. 2) Appreciation for contextually-embedded problem-solving. Important issues arise when well-intentioned systems evolve post-deployment. The NRT emphasizes context early and often as research is being formulated. Summer bootcamps will facilitate research problem formulation that enables early cycles of feedback and testing with a broad set of stakeholders. Additionally, by intertwining students from different programs by engaging them in a professional certificate through the onboarding sequences, informal opportunities will be created for natural cross-pollination from theory to practice and back. 3) Holistic professional identities. Although many Ph.D. programs are starting to build scaffolding to support “soft-skills,” this usually occurs separately from core program elements. This program’s strategy is to make communication, leadership, teamwork, and ethics central to each program element. The bootcamps and seminars will also provide structured opportunities for students to learn, practice, and reinforce their strategies, e.g., engaging with ethics in context. 4) Valuing diverse perspectives in decision-making. AI algorithms tend to exacerbate existing biases, making it especially important to bring diverse perspectives into decision-making to mitigate unintended consequences. Currently, AI adoption is being driven by a relatively homogenous group. There is a need to increase participation from underrepresented groups and expose students to the value of bringing in diverse perspectives early in the process. The NSF Research Traineeship (NRT) Program is designed to encourage the development and implementation of bold, new potentially transformative models for STEM graduate education training. The program is dedicated to effective training of STEM graduate students in high priority interdisciplinary or convergent research areas through comprehensive traineeship models that are innovative, evidence-based, and aligned with changing workforce and research needs.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.

随着人工智能（AI）越来越深入地融入现代工作场所的结构，工作的性质和结构正在发生根本性的变化。这种整合在无处不在的人工智能改变工作场所的机会与围绕偏见，安全和隐私的新兴风险之间产生了紧张关系。目前，人工智能工具正在以异常快的速度开发，并部署到价值最大化先于监管的环境中。因此，下一代创新者需要一种新的培训。对于算法设计者来说，这意味着理解并敏感于他们的创作可能通过与社会技术生态系统中的用户交互以意想不到的方式运行的上下文。对于系统设计师来说，这意味着要充分了解人工智能工具是如何发展的，以重新设想任务和流程如何能够并且应该以充分利用人工智能工具潜在力量的方式改变工作。这个国家科学基金会研究培训（NRT）奖给乔治华盛顿大学将通过培训博士生，硕士生和研究生证书的学生来解决这些需求，他们将准备在未来的工作场所以积极影响社会的方式为人工智能做出融合的研究贡献。该项目预计将培训一百二十（120）名学生，其中包括二十五（25）个资助的博士学位。该NRT主要为计算机科学和系统工程专业的学生提供服务，但与法律，媒体，公共事务，公共卫生和国际事务的学生和教师密切互动。该NRT旨在培养能够“共同设计”AI算法和工作系统的研究人员，以解锁新系统的能力及其“可信度”的新机会。为了实现这一目标，教育计划的目的是灌输以下内容：1）在桥接遥远的学科舒适。通过新颖的入职序列和与同行，导师和行业的跨学科参与的共享经验，该计划将教育跨学科的“梳形”学者，他们在人工智能算法或工作系统设计方面具有坚实的基础，并且也很舒服深入参与其他学科领域对他们选择的研究问题至关重要。2)对情境嵌入式问题解决的欣赏。当意图良好的系统在部署后发展时，就会出现重要的问题。NRT强调早期背景，并经常在研究制定过程中。夏令营将促进研究问题的制定，使早期的反馈周期和测试与广泛的利益相关者。此外，通过让学生通过入职序列获得专业证书，将来自不同课程的学生交织在一起，将为从理论到实践再回来的自然交叉授粉创造非正式的机会。3)完整的专业身份。虽然很多博士程序开始构建脚手架来支持“软技能”，这通常与核心程序元素分开发生。该计划的战略是使沟通，领导，团队合作和道德中心的每个程序元素。训练营和研讨会还将为学生提供学习，实践和加强他们的策略的结构化机会，例如，与道德结合起来。4)在决策中重视不同的观点。人工智能算法往往会加剧现有的偏见，因此将不同的观点纳入决策过程以减轻意外后果尤为重要。目前，人工智能的采用是由一个相对同质的群体推动的。有必要增加代表性不足的群体的参与，并让学生了解在这一进程的早期引入不同观点的价值。 NSF研究培训（NRT）计划旨在鼓励为STEM研究生教育培训开发和实施大胆的，新的潜在变革模式。该计划致力于通过创新的、基于证据的、与不断变化的劳动力和研究需求相一致的综合培训模式，在高优先级的跨学科或融合研究领域对STEM研究生进行有效培训。该奖项反映了NSF的法定使命，并通过使用基金会的智力价值和更广泛的影响审查标准进行评估，被认为值得支持。

项目成果

DP2-Pub: Differentially Private High-Dimensional Data Publication With Invariant Post Randomization

• DOI: 10.1109/tkde.2023.3265605
• 发表时间: 2022-08
• 期刊: IEEE Transactions on Knowledge and Data Engineering
• 影响因子: 8.9
• 作者: Honglu Jiang;Hao-Chun Yu;Xiuzhen Cheng;Jian Pei;Robert Pless;Jiguo Yu

Emotion and Virality of Food Safety Risk Communication Messages on Social Media

• DOI: 10.4148/1051-0834.2391
• 发表时间: 2021-09
• 期刊: Journal of Applied Communications
• 作者: X. Wang;Xiaoli Nan;S. Stanley;Yuan Wang;L. Waks;David A. Broniatowski

The Opportunists in Innovation Contests: Understanding Whom to Attract and How to Attract Them

创新竞赛中的机会主义者：了解吸引谁以及如何吸引他们

• DOI: 10.1080/08956308.2022.2132771
• 发表时间: 2023
• 期刊: Research-Technology Management
• 影响因子: 2.2
• 作者: Vrolijk, Ademir;Szajnfarber, Zoe
• 通讯作者: Szajnfarber, Zoe

Understanding Post-Production Change and Its Implications for System Design: A Case Study in Close Air Support During Desert Storm

了解后期制作变化及其对系统设计的影响：沙漠风暴期间近距离空中支援案例研究

• 发表时间: 2022
• 期刊: Naval engineers journal
• 影响因子: 0.2
• 作者: Singh, Aditya;Szajnfarber, Zoe
• 通讯作者: Szajnfarber, Zoe

Knowledge-Augmented Language Models for Cause-Effect Relation Classification

• DOI: 10.18653/v1/2022.csrr-1.6
• 发表时间: 2021-12
• 期刊: Proceedings of the First Workshop on Commonsense Representation and Reasoning (CSRR 2022)
• 作者: Pedram Hosseini;David A. Broniatowski;Mona T. Diab