Conducting Participatory Design to Improve Algorithms in Public Services: Lessons and Challenges

Abstract

Government agencies are increasingly looking towards algorithmic decision-making systems as a means to reduce costs and optimize processes. However, these algorithms are being constructed in an opaque and isolated manner with calls to adopt a more participatory approach such that stakeholders become co-designers in the process. We share our experiences from conducting participatory design to improve algorithms in the Child-Welfare System. We discuss a policy-mandated algorithm and an agency-level theory-driven algorithm to show how tensions arise when the values of workers are not embedded in the design of an algorithm.