Child Welfare System: Interaction of Policy, Practice and Algorithms

This paper focuses on understanding the collaborative work of multi-disciplinary teams in the child welfare system (CWS). CWS workers participate in meetings mediated by policies in place, current child-welfare practice, as well as algorithms that offer recommendations. We conducted 25 observations of these meetings to assess how algorithms aid decision-making in a domain where decisions often come down to the policies and practices in place. Our findings suggest that the algorithm works fairly well at recommending placement settings, however, these recommendations are often overridden because of policy or legal requirements. Moreover, re-appropriation of the placement algorithm to prescribe the rates for foster parents has led to unintended consequences. This poster identifies uses cases of the algorithm in place, scenarios where conflicts arise between the algorithm and policy/practice, as well as how these conflicts are addressed. Our work identifies a need for human-centered algorithms that can better support child welfare practice.