The reports appear with distressing regularity. Each describes yet another child maltreated in New Mexico, even while being monitored by the state’s Children, Youth and Families Department. These reports barely outnumber those critical of CYFD itself, of its procurement practices, its outdated computer systems and its lack of transparency.
And don’t forget the reports of problems within the state’s treatment foster care agencies. New Mexico contracts with these agencies to coordinate therapeutic care for high-needs children in state custody. But like CYFD itself, several of these agencies have been plagued by administrative problems. Some have lost their operational licenses and others face lawsuits over their handling of cases.
Of course, there are instances when our state’s child welfare system works. I’ve met foster parents in New Mexico who are exemplars, who try to work within the system to give at-risk kids the care they need. Nevertheless, even these experienced foster care providers tell of having to repeatedly navigate nonsensical bureaucratic hurdles to maintain a home for the children under their care. Yet the same hurdles have failed to prevent the abuse and deaths of other children.
The CYFD system has seemingly not found the right balance among competing goals of reuniting families, protecting vulnerable children, and reducing the bureaucratic burden on successful foster care families. Unfortunately, finding the right balance may be impractical because CYFD employees are charged with an unenviable and difficult task: predicting the future. What would be best for a child, particularly in light of continually changing circumstances?
To better predict the future and rebalance the system, there is an option that could help: Remove the human element from some CYFD decision-making processes. Instead, where appropriate and feasible, CYFD should consider using methodologies called “predictive analytics” to improve child welfare outcomes. Predictive analytics is a broad term for advanced statistical techniques that use large quantities of data to predict the likelihood of future outcomes, including child welfare outcomes such as repeat maltreatment or foster care re-entry.
Predictive analytics can supplement good casework and administration by providing data-driven insights into the relative risks children face amid complex and varying conditions. It’s not a panacea for all of CYFD’s challenges. But it could serve as an important tool to support agency policies and processes and the employees themselves.
Some advocates for children or families might be concerned about permitting statistical algorithms to play a role in life-changing decisions. However, a growing body of research is demonstrating that in forecasting outcomes, including the risk assessment that’s integral to child welfare decisions, these algorithms consistently outperform humans. Predictive analytics can overcome many of the biases and flaws inherent in human decision-making.
CYFD already uses some risk assessment tools in its work. But it needs a better system for developing predictions specific to child welfare outcomes. CYFD would not be starting from scratch. The federal government has completed preliminary work in this area and offers guidance to states on how to implement predictive analytics in child welfare systems.
Currently, CYFD appears to be trying versions of past attempts at making improvements – new leadership, more evaluations, different contracts for services. But like the rest of us, the humans involved in our state’s child welfare system are all-too fallible. Perhaps it’s time to try a new approach:
Let an algorithm guide the humans.
Shana Judge, JD, PhD, is a social science researcher and data scientist. She is the founder and Executive Director of Duddon Evidence to Policy Research, an Albuquerque-based law and policy consulting business – duddonresearch.org.
