FW-HTF-P: Combining Human and Machine Intelligence to Improve Equity and Fairness in the Work of Municipal, Public Sector Decision Making

政府机构和公共部门越来越多地转向数据驱动和基于人工智能（AI）的技术，以提高人类决策和服务提供的质量。这种向人工智能驱动方法的转变有望为人手不足的部门提供急需的支持，并加快满足社区成员关注的速度。然而，过度强调效率和有效性，沿着大多数基于人工智能的系统对如何提出特定建议缺乏透明度，使得公共部门工作人员难以做出公平，公正，有效的决策。这可能导致缺乏对选民的问责，并对少数群体社区造成伤害，如预测性警务系统中的种族偏见等案例所示。该项目将使用服务设计启发的方法来调查公共部门数据和算法基础设施是如何形成的，可能导致不同市政服务提供的因素，以及如何重新设计这些基础设施以解决它们有时会加剧的不公平现象。这项规划补助金将使项目团队对问题有更多的了解，同时帮助团队建立社区伙伴关系并扩大其学科专业知识，以制定更大的长期研究议程。该项目涉及四项主要活动，这些活动将在企业主与使用或计划使用数据创新和人工智能增强决策工具的公共部门许可证，许可证和检查部门的互动中进行。第一项活动涉及使用数据审计方法描述收集和分析的数据的当前状态，目的是确定所需的变化，以按与公平和公正有关的质量对企业所有者进行分类，并支持业务流程挖掘。第二项和第三项活动是了解公共部门工作人员的优先事项和经验，这些工作人员决定许可证、执照和检查，以及需要这些服务以启动企业的地方、小型、妇女和少数民族拥有的企业所面临的经验和挑战。这些活动将以与相关人群成员的访谈为基础，最后将开展第四项活动，在该活动中，这些利益攸关方群体将与研究小组一起参加共同设计讲习班，以帮助确定和平衡所有利益攸关方的关键目标，沿着关于如何塑造系统以实现人工智能中的这些目标的想法-该奖项反映了NSF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。

Government agencies and the public sector have increasingly turned to data-driven and Artificial Intelligence (AI)-based techniques to enhance the quality of human decision-making and service provision. This shift toward an AI-driven approach holds promise in its ability to offer much-needed support to understaffed departments and increasing the speed with which community members’ concerns are met. However, an over-emphasis on efficiency and effectiveness, along with most AI-based systems’ lack of transparency into how a given recommendation is made, have made it difficult for public sector workers to make fair, equitable, efficient decisions. This can result in lack of accountability to constituents and cause harm to minoritized communities, as demonstrated by cases such as racial bias in predictive policing systems. This project will use a service design-inspired approach to investigate how public sector data and algorithmic infrastructures are formed, factors that can lead to disparate municipal service provision, and how these infrastructures might be redesigned to address the inequities they sometimes reinforce. This planning grant will give the project team additional insight into the problem while helping the team build community partnerships and expand its disciplinary expertise, in order to develop a larger, long-term research agenda. The project involves four major activities, which will be conducted in the context of business owners’ interaction with public sector permit, license, and inspection departments that use or plan to use data-innovation and AI-enhanced decision-making tools. The first activity involves characterizing the current state of data collected and analyzed using a data audit methodology, with the goal of identifying needed changes to disaggregate business owners by qualities related to equity and fairness and to support business process mining. The second and third activities are to understand the priorities and experiences of public sector workers who make decisions about permits, licenses, and inspections, and the experience and challenges faced by local, small, women- and minority-owned businesses that need these services in order to launch their ventures. Those activities will be based on interviews with members of the respective populations, culminating in a fourth activity in which these stakeholder groups will come together with the research team in co-design workshops to help define and balance the key objectives held by all stakeholders, along with ideas for how systems can be shaped to achieve those objectives in AI-based systems that support public sector 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.