Original
Navigating the Algorithmic State: Public Values and Citizens' Preferences of AI ‐Driven Public Service Delivery
Abstract
ABSTRACT As the public sector increasingly adopts AI‐powered automated decision systems (ADS), understanding how citizens experience and value ADS use in public decision‐making is both normatively and practically important. Therefore, we examine and compare the effects of seven attributes of public values on citizens' support for ADS adoption in two real‐world policy areas: healthcare and criminal justice. The results of an online conjoint experiment with approximately 1500 American adults reveal that effectiveness, transparency, and accountability are the most salient values influencing citizens' acceptance of government ADS deployment. Subgroup analyses suggest that algorithmic equity and data protection are significantly prioritized in the enforcement‐oriented criminal risk score prediction scenario, relative to the service‐oriented healthcare subsidy area. These findings provide insights into how to integrate public values in the design of government ADS that reflect citizen input and preferences.
中文
驾驭算法国家：公共价值与公民对人工智能驱动公共服务供给的偏好
摘要
摘要：随着公共部门日益采用人工智能驱动的自动化决策系统（ADS），理解公民如何体验和评价ADS在公共决策中的使用，无论在规范层面还是实践层面都十分重要。因此，我们考察并比较公共价值的七项属性对公民支持在两个现实政策领域——医疗保健和刑事司法——采用ADS的影响。一项约1500名美国成年人参与的在线联合实验结果显示，有效性、透明度和问责制是影响公民接受政府部署ADS的最显著价值。亚组分析表明，相对于服务导向的医疗补贴领域，在执法导向的犯罪风险评分预测情境中，算法公平和数据保护显著更受优先考虑。这些发现为如何在政府ADS设计中整合公共价值以反映公民意见和偏好提供了启示。
关键词
人工智能、自动化决策系统、公共价值、公共服务供给、公民偏好、联合实验、算法治理、政府技术采纳