Automating Social Security in the UK: A Study on Incorporating Claimant Voices in the Design of Universal Credit

世界各国政府都在采用算法系统来提供基本的社会服务，从警务到欺诈侦查再到儿童福利。虽然越来越多的研究调查了这些系统的设计——它们潜在的政策目标和有效性——但在自动化如何影响用户以及用户的观点如何民主地告知这些系统的技术、法律和行政设计方面，文献中存在空白。该项目旨在通过对英国通用信贷计划的研究，更广泛地了解自动化社会服务的出现及其对索赔人的影响。拟议的项目将通过两组研究问题来检查英国社会保障福利的自动化。(1)什么政治价值观和政策推动了构成通用信用的技术系统？通过对DWP工作人员的访谈和文件分析，该项目将阐明通用信贷更广泛的政治背景，并成为第一批绘制该系统技术维度的学术研究之一。(2)索赔人如何与通用信贷的自动化组件互动，社会保障中的自动化是否会损害某些人群？通过与与索赔人合作的非政府组织的访谈和与索赔人自己的研讨会，我将研究通用信贷的自动化功能是否以及如何嵌入某些对人们产生不同影响的偏见或伤害。用户讲习班将探讨问责制过程，鼓励索赔人对系统本身的设计提供反馈；这些活动将为与用户进行基于证据的磋商提供模板，以形成算法社会服务。研究将采用多种方法，包括文件分析、实地考察、访谈和用户设计讲习班：(a)文件分析将揭示通用信贷更广泛的政治背景和发展历史。该项目将利用公共文件，如立法和法令、部门报告、网站、指导方针、议会委员会报告和审计，以及公共记录请求，以获取系统的技术规范，如供应商合同和特定自动化功能的设计。(b)访谈将进一步阐明UC的设计及其对索赔人的不同影响，包括与DWP工作人员进行约15次访谈，接触DWP UC用户研究团队的项目合作伙伴，以及通用信贷数字服务团队和DWP的福利和养老金数字技术服务；并与与索赔人合作或代表索赔人的苏格兰非政府组织的工作人员进行了25次访谈。c)通过在爱丁堡地区当地的公民咨询局（Citizen Advice Bureau）实习，我将在12个月内每周进行一天的实地考察，以观察Universal Credit与系统自动化功能相关的问题出现的频率。d)最后，我将举办六个设计研讨会，重点关注苏格兰背景下被数字排除的索赔人样本，DWP的UC用户研究团队发现这一群体特别难以进行用户测试，以及对其数字技能充满信心的索赔人比较组。UC中的自动化是一个复杂的主题，跨越了多个知识领域，因此需要一个综合的方法，结合政治，设计和自动化社会保障监督的多个观点。该研究特别解决了缺乏关于索赔人与自动化系统交互的经验的文献。通过公开报道、公共活动和媒体传播，该项目将引发对索赔人在使用这些系统时面临的问题的讨论，并探索如何将索赔人的声音融入到他们未来的修改中。

Governments around the world are adopting algorithmic systems to deliver fundamental social services, from policing to fraud detection to child welfare. While a growing set of research investigates the designs of these systems - their underlying policy goals and effectiveness - there is a gap in the literature on how automation impacts users and how user perspectives could democratically inform these system's technical, legal and administrative design. This project sets out to achieve a wider understanding of the emergence of automated social services and their effects on claimants through a study of the UK's Universal Credit scheme. The proposed project will examine the automation of social security benefits in the U.K. through two sets of research questions.(1) What political values and policies drive the technical systems that comprise Universal Credit? By interviewing DWP staff and through document analysis, the project will illuminate Universal Credit's wider political context and be among the first academic studies to map the system's technical dimensions. (2) How do claimants interact with automated components of Universal Credit and does automation in social security harm certain populations? Through interviews with NGOs that work with claimants and workshops with claimants themselves, I will examine whether and how automated features of Universal Credit may embed certain biases or harms that impact people differentially. The user workshops will explore accountability processes that encourage claimants to offer feedback on the design of the system itself; these events will provide templates for evidence-based consultations with users to shape algorithmic social services. The research will involve mixed methods, including document analysis, fieldwork, interviews, and user design workshops:(a) Document analysis will reveal Universal Credit's wider political context and history of development. The project will draw on public documents such as legislation and decrees, departmental reports, websites, guidelines, parliamentary committee reports, and audits, as well as public records requests to gain access to technical specifications of the system, such as vendor contracts and designs of specific automated functions. (b) Interviews will further illuminate the design of UC and its differential impacts on claimants, including approximately 15 interviews with DWP staff, reaching out to project partners on the DWP's UC User Research team, as well as the Universal Credit digital services team and the DWP's Benefits and Pensions Digital Technology Services; and 25 interviews with staff at Scottish NGOs that work with or on behalf of claimants. c) Through placement in a Citizen Advice Bureau local to the Edinburgh region, I will conduct fieldwork one day a week for 12 months to observe how often problems with Universal Credit relate to automated features of the system. d) Finally, I will hold six design workshops, focusing on a sample of digitally excluded claimants in the Scottish context, a group that the DWP's UC User Research team has found particularly difficult to reach for user testing, as well as a comparison group of claimants who are confident in their digital skills.Automation in UC is a complex topic that spans several knowledge domains and therefore demands an integrated approach, combining multiple perspectives on the politics, designs, and oversight of automated social security. The study particularly addresses a lack of literature on the experiences of claimants as they interact with automated systems. Through public reports, public events and media dissemination, the project will spark discussion on the problems claimants face while using these systems and explore how claimant voices can feed into their future modifications.