AI in Social Services: Ensuring Fairness and Preventing Bias

AI in Social Services: Protecting Vulnerable Populations

As artificial intelligence becomes more integrated into social service functions like child and elder abuse investigations, concerns are being raised about potential biases against already disadvantaged groups. A recent study by Aislinn Conrad, an associate professor at the University of Iowa and former social worker, highlights the need for strict safeguards.

Conrad’s research indicates that while AI can efficiently process vast amounts of data to pinpoint individuals at risk, it risks overlooking the critical human judgment essential for making fair decisions. This is particularly crucial when decisions could disrupt or marginalize families.

Recommendations for AI Use in Social Services

Conrad suggests that social service agencies should prioritize survivors’ needs when deploying AI systems. Her key recommendations include:

Prioritize Human Input: Families should have a say in AI usage, understand algorithmic decisions, and possess the right to challenge them.

Audit for Fairness: AI systems must undergo regular reviews to prevent biased or discriminatory results.

Limit Surveillance: Use AI to connect families with essential resources rather than increase monitoring.

Why It Matters

AI’s growing role in decision-making can significantly impact children and survivors of abuse. Without careful oversight, these systems might perpetuate existing biases, disproportionately affecting those facing social or economic hardships. The study offers practical methods to implement AI responsibly by ensuring human involvement, checking for bias, and focusing on resource allocation.

Conrad emphasizes, “AI can make us more efficient and effective, but only if we’re willing to do the hard work of developing responsive systems that keep families safe without reproducing bias.”

Conrad’s experience in child protective services, particularly in rural Kansas areas affected by drug use and poverty, has informed her research. She questions the link between poverty and abuse, wondering if it stems from increased surveillance rather than a higher incidence of abuse.

In Iowa, these findings are particularly relevant as the state’s Department of Health and Human Services collaborates with Google on VISION, an AI-driven system aimed at modernizing child welfare case management. Conrad warns that such systems might misinterpret data, such as housing instability or public assistance, as risk factors, increasing scrutiny on already struggling families.

“When historical data shows that poor families and families of color have been disproportionately surveilled by child welfare systems, the algorithm learns that pattern and repeats it at scale, and does so invisibly,” Conrad points out. “A ‘neutral’ algorithm trained on biased data is just biased automation.”

However, AI could also help identify communities needing resources due to severe housing instability or inadequate services. The ultimate aim, according to Conrad, should be improving access to services rather than increasing monitoring.

The study, titled “Centering survivor agency in AI for violence prevention,” was published online on October 9 in the Journal of the Society for Social Work and Research. Conrad is the sole author, and her research was self-funded.