FAI: Toward Fair Decision Making and Resource Allocation with Application to AI-Assisted Graduate Admission and Degree Completion

FAI：通过应用于人工智能辅助研究生入学和学位完成来实现公平决策和资源分配

• 批准号: 2147276
• 负责人: Furong Huang
• 金额: $ 62.5万
• 依托单位: University of Maryland, College Park
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2022
• 资助国家: 美国
• 起止时间: 2022-02-15 至 2025-01-31
• 项目状态: 未结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2147276&HistoricalAwards=false
• 关键词: FAI; Toward; Fair; Decision; Making

Machine learning systems have become prominent in many applications in everyday life, such as healthcare, finance, hiring, and education. These systems are intended to improve upon human decision-making by finding patterns in massive amounts of data, beyond what can be intuited by humans. However, it has been demonstrated that these systems learn and propagate similar biases present in human decision-making. This project aims to develop general theory and techniques on fairness in AI, with applications to improving retention and graduation rates of under-represented groups in STEM graduate programs. Recent research has shown that simply focusing on admission rates is not sufficient to improve graduation rates. This project is envisioned to go beyond designing "fair classifiers" such as fair graduate admission that satisfy a static fairness notion in a single moment in time, and designs AI systems that make decisions over a period of time with the goal of ensuring overall long-term fair outcomes at the completion of a process. The use of data-driven AI solutions can allow the detection of patterns missed by humans, to empower targeted intervention and fair resource allocation over the course of an extended period of time. The research from this project will contribute to reducing bias in the admissions process and improving completion rates in graduate programs as well as fair decision-making in general applications of machine learning.This project will focus on machine learning algorithms for resource allocation, which can be used at various points throughout a process such as in education. The team will propose new notions of fairness and show the applicability of those notions to settings in which limited resources, such as acceptance to the program, faculty mentoring, professional development, and paid assistantships or fellowships, are allocated to students fairly. The proposed research will also go beyond fairness in task-specific supervised learning settings and investigate fairness in unsupervised learning that guarantees to learn fair representations or generative models for multiple downstream tasks. The team will address the practical problems that arise due to uncongenial data in real-world sequential decision-making systems, including distribution shifts between training and test, imbalanced data, and missing sensitive attributes. This proposal contains a comprehensive plan to incorporate its research into education at high school, undergraduate, and graduate levels, as well as plans for within- and cross-disciplinary dissemination of research results, outreach, and other synergistic activities.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.

机器学习系统在日常生活中的许多应用中变得突出，例如医疗保健，金融，招聘和教育。这些系统旨在通过发现大量数据的模式来改善人类决策，而不是人类可以直觉的。但是，已经证明这些系统学习并传播了人类决策中存在的类似偏见。该项目旨在开发有关AI公平性的一般理论和技术，并应用于提高STEM研究生计划中代表性不足群体的保留率和毕业率。最近的研究表明，仅仅关注入学率不足以提高毕业率。设想该项目超越设计“公平的分类器”，例如公平的研究生承认，可以在一个时间内满足静态公平的概念，并设计AI系统，这些系统在一段时间内做出决定，目的是确保整体长期公平的公平成果。数据驱动的AI解决方案的使用可以使人们无法检测人类所遗漏的模式，从而在长时间的时间内赋予有针对性的干预和公平资源分配。 该项目的研究将有助于减少录取过程中的偏见，并提高研究生课程的完成率，以及在机器学习的一般应用中的公平决策。该项目将重点介绍用于资源分配的机器学习算法，这些算法可以在整个教育中的各个过程中使用。该团队将提出新的公平概念，并将这些概念在设置中的适用性，其中有限的资源（例如接受该计划，教师指导，专业发展以及有偿助手职位或奖学金）的资源公平分配给学生。拟议的研究还将超越特定于任务的监督学习环境中的公平性，并调查无监督学习中的公平性，保证为多个下游任务学习公平表示或生成模型。该团队将解决由于现实世界中的顺序决策系统中不兼容的数据而引起的实际问题，包括培训和测试之间的分配变化，数据不平衡的数据以及缺失的敏感属性。该提议包含一个全面的计划，将其在高中，本科和研究生水平的教育中纳入研究，以及在研究成果，外展和其他协同活动的内部和跨学科传播计划中。该奖项反映了NSF的法规使命，并认为通过基金会的知识优点和广泛的评估，可以通过评估来进行评估。

项目成果

Rethinking Adversarial Policies: A Generalized Attack Formulation and Provable Defense in RL

• 发表时间: 2023-05
• 作者: Xiangyu Liu;Souradip Chakraborty;Yanchao Sun;Furong Huang

Secure Sampling with Sublinear Communication

使用次线性通信进行安全采样

• 发表时间: 2022
• 期刊: Springer
• 作者: Choi, Seung Geol;Dachman-Soled, Dana;Gordon, S. Dov;Liu, Linsheng;Yerukhimovich, Arkady
• 通讯作者: Yerukhimovich, Arkady

Live in the Moment: Learning Dynamics Model Adapted to Evolving Policy

• DOI: 10.48550/arxiv.2207.12141
• 发表时间: 2022-07
• 期刊: ArXiv
• 作者: Xiyao Wang;Wichayaporn Wongkamjan;Furong Huang

Exploring and Exploiting Decision Boundary Dynamics for Adversarial Robustness

• DOI: 10.48550/arxiv.2302.03015
• 发表时间: 2023-02
• 期刊: ArXiv
• 作者: Yuancheng Xu;Yanchao Sun;Micah Goldblum;T. Goldstein;Furong Huang

Game-Theoretic Robust Reinforcement Learning Handles Temporally-Coupled Perturbations

• DOI: 10.48550/arxiv.2307.12062
• 发表时间: 2023
• 期刊: ArXiv
• 作者: Yongyuan Liang;Yanchao Sun;Ruijie Zheng;Xiangyu Liu;T. Sandholm;Furong Huang;S. McAleer