EAGER: Perceptions of Fairness and Justice in AI Software for Talent Acquisition

在工作招聘和雇佣中感受到的公平和正义受到几个因素的影响。一些因素包括人员和时间决策过程的一致性、及时和信息丰富的反馈、面试问题的适当性以及就业前测试似乎与工作要求有关的程度。这些因素共同影响着招聘和聘用的决定，而且越来越多地在人工智能(AI)的帮助下做出决定。在这个项目中，一个社会技术框架被应用于探索人工智能支持的人才获取算法的公平和正义的感知。调查员将引出并分析人力资源人员、非裔美国求职者和人工智能软件设计师的看法。这些结果将被用来为人类和正在使用的算法的偏见识别和缓解程序和技术的设计提供参考。这项探索性研究的智力价值在于开发了可以用来衡量算法公平和正义感知的定性工具和度量标准。该研究方法通过三管齐下的方法扩展了遴选制度感知公平的程序规则理论，包括IT行业代表性不足的求职者、管理人才获取过程的人力资源专业人员以及设计以公平为产品设计和开发核心价值的人工智能软件的IT专业人员。考察了对情景的认知，以及受这些决定影响的求职者的实际经历。这项研究有助于评估算法的公平性，因为目前几乎没有人深入了解历史上被边缘化的人群可能会如何看待人工智能系统或受到人工智能系统的不利影响。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。

Perceived fairness and justice in job recruiting and hiring are influenced by several factors. Some factors are the consistency of the decision-making process across people and time, timely and informative feedback, propriety of the interview questions, and the extent to which pre-employment tests appear to relate to the job requirements. These factors come together to influence decisions about recruiting and hiring and are being made increasingly with the help of artificial intelligence (AI). In this project, a sociotechnical frame is applied to explore perceptions of fairness and justice of AI-supported talent acquisition algorithms. the investigator will elicit and analyze perceptions of human resources personnel, African American job seekers, and AI software designers. The outcomes will be used to inform the design of bias recognition and mitigation procedures and technologies for both humans and the algorithms being used.The intellectual merit of this exploratory study is the development of qualitative instruments and metrics that can be used to measure perceptions of algorithmic fairness and justice. The research approach extends a theory of procedural rules for perceived fairness of selection systems by using a three-pronged approach comprising job seekers who are under-represented in the IT industry, human resource professionals who manage the talent acquisition process, and IT professionals who design AI software with fairness as the core value in product design and development. Perceptions using scenarios are examined as well as the actual experiences of jobseekers who are affected by these decisions. This research contributes to an assessment of algorithmic fairness at a time when there is currently little insight into how historically marginalized populations might perceive or be adversely affected by AI systems.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.