Predictive analytics in child welfare: An overview of current initiatives and ethical frameworks to inform equitable policy and practice

Paul Lanier, Phd
University of North Carolina at Chapel Hill

Daniel Gibbs, JD
University of North Carolina at Chapel Hill

ABSTRACT:
The child welfare system is faced with high-stakes decisions every day that must balance goals of child protection with parental rights. The use of administrative data coupled with advanced analytic methods to drive decision-making in child welfare practice and policy has increased in recent years. For example, predictive analytics and associated approaches have been used to identify families at higher risk for future harm to target appropriate services. However, the potential for bias in algorithmic decision-making must be weighed against the potential benefits of using any new technology. Specifically, we must consider whether predictive analytics exacerbates or ameliorates existing racial differences in experiences with the child welfare system. In this presentation, we will 1) provide an overview of the current state of the field, 2) give several current examples of predictive analytics used in child welfare, and 3) present preliminary findings from our recent study that aimed to apply an existing ethical framework to a data-driven prevention initiative known as Birth Match.