DASS: Exploring how convergence methods foster shared accountability to reveal, map, and mitigate the sources and dynamics of bias across social service provisioning systems

传统上，社会服务是围绕其使命组织的，如教育、安全或健康。一种较新的方法，称为“包裹式服务”或“护理系统”，围绕个人及其特定的背景和需求来组织服务。这些系统在应用于现实世界时面临许多挑战。应用程序流程通常更多地关注技术的潜力，而较少关注社区的现实、历史和需求。拟议的研究通过评估护理系统在现实世界中的实施情况来解决这一差距。这项研究涉及研究如何更好地支持公民参与，并在社区护理系统的发展和整合中减少偏见。这项提议的一个关键成果是了解公正、公平、多样性和包容性所激发的程序的影响，以有目的地检查公共服务提供软件系统的设计和应用中的偏见和减少差异。拟议工作的目标是通过开发可概括和可本地化的实践来推进可问责的软件系统，以探索如何识别特定和系统的偏见来源，改善公共服务提供结果，并将与有偏见的计划结果的差异降至最低。拟议工作建立在现有理论和项目团队在设计和执行支持公众服务的软件系统方面的经验基础上。其目的是确定和绘制不同类型的对责任的威胁，这些威胁可以在护理服务的社会计算生态系统中预见。该团队正在研究偏见如何通过法规、社会背景、数据以及软件开发过程中缺乏代表性等因素在整个系统中传播的动态。一旦这种动态被绘制出来，该团队将探索如何通过与社区成员合作计划和实施的一系列干预措施来关注偏见。这些将包括协作绘制系统图，举办知识汇聚研讨会、算法影响评估，以及使用可在设计和部署这些护理系统时增加治理机会的计算分析技术。在这一过程中要跟踪的一个结果衡量标准是，一个系统在应对个人一级的威胁和通过共同管理和问责在系统一级创造偏见豁免权方面的有效性。这项研究预期的产品类型是理论上的进步(阐明新的威胁，对动力学和治疗的新理解)，与算法责任政策工具包相关的特定产品，以及通过可概括的融合工作坊根据本地用例定制的实践指南。该奖项反映了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。

Social services, traditionally, have been organized around their missions, such as education or safety or health. A newer approach, called "wrap-around services" or "systems of care," organizes services around individuals and their specific context and needs. These systems face many challenges when applied in real-world settings. Application processes often focus more on the potential of technologies and less on the realities, histories, and needs of communities. The proposed research addresses this gap by evaluating the implementation of a system of care in a real-world setting. The research involves studying how civic participation may be better supported and bias reduced in the development and integration of systems of care for communities. A key outcome of this proposal is to understand the influence of processes inspired by justice, equity, diversity, and inclusion to intentionally check bias and reduce disparities in the design and application of public service provisioning software systems. The goal of the proposed work is to advance accountable software systems through developing generalizable and localizable practices for exploring how to identify specific and systemic sources of bias, improve public service provisioning outcomes, and minimize disparities from biased program outcomes.The proposed work builds from existing theory and the project team’s experience in designing and executing software systems that support services to the public. The intent is to identify and map the different types of threats to accountability that can be anticipated within the socio-computational ecosystem of care services. The team is studying dynamics of how bias might propagate across the system through regulations, social contexts, data, and lack of representation in the software development process, among other factors. Once such dynamics are mapped, the team will explore how to attend to bias through a series of interventions planned and conducted in collaboration with community members. These will include mapping the system collaboratively, and holding knowledge convergence workshops, algorithmic impact assessments, and using computational analytic techniques that can augment governance opportunities in the design and deployment of these systems of care. An outcome measure to be tracked through this process is the effectiveness of a system at responding to individual-level threats and creating bias immunity at the systems level through shared stewardship and accountability. The type of products expected from this research are advances in theory (articulation of novel threats, new understandings of dynamics & treatments), specific products related to an algorithmic accountability policy toolkit, and guidelines for practice customizable to local use cases through generalizable convergence workshops.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.