Automated Decision Systems: Child Welfare Predictive Analytics Tools

By ACLU of Washington.

Summary: Investigates the use of predictive-risk algorithms in child welfare decision-making and their implications for families.

Relevance: Directly relevant to the Children, Youth & Families subdomain: examines how predictive-risk tools used in CPS decision-making actually predict removal likelihood rather than maltreatment itself, with concrete stakes for social workers navigating child welfare systems on behalf of families.

Key findings: In Los Angeles County, a rejected model correctly flagged 76% of severe-harm cases but also flagged 3,829 additional cases as high-risk, 95% of which did not involve such harm; Allegheny County’s screening tool did not accelerate the pre-existing decline in screen-ins for Black and white children and, per the piece, “largely halted” that decline after adoption through 2021.

Read the original: https://www.aclu-wa.org/news/automated-decision-systems-child-welfare-predictive-analytics-tools/