AI in Social Services Needs Bias Audits and Human Oversight, Researcher Argues
A University of Iowa social work researcher warns that artificial intelligence used in child abuse, intimate partner violence, and elder abuse cases could introduce bias against vulnerable groups. The study recommends giving survivors meaningful input, auditing systems for fairness, and using AI to connect people with resources rather than expand surveillance. It argues that automation should not replace the human judgment needed for fair decisions.
The study appears in the Journal of the Society for Social Work and Research. Its author, Aislinn Conrad, is an associate professor at the University of Iowa School of Social Work and previously investigated child protective cases and helped reunite foster children with families in rural Kansas.
Iowa's Department of Health and Human Services partnered with Google in February to build VISION, an AI system for child welfare case and data management. Conrad cautions that indicators such as public aid, unstable housing, or earlier agency contact might be treated as risk signals, potentially increasing scrutiny of families already facing hardship.
If adopted, these safeguards could shape how agencies use AI in child abuse, intimate partner violence, and elder abuse cases. Families and survivors may gain clearer ways to question automated flags, while caseworkers could see changes in workload and discretion. Fairness audits and resource-focused tools might reduce harms, but outcomes depend on implementation, oversight, and whether affected people have meaningful voice.