Child Welfare and Predictive Analytics: Safety in Numbers?

When Beverly Walker accepted her job as director of the Illinois Department of Children and Family Services (DCFS), she knew she was walking into a warzone. Hammered by allegations of mismanagement and incompetence, the agency had seen nine directors and acting directors in six years, and the most recent director had resigned amid an ethics probe and a series of high-profile child fatalities.
Walker likened DCFS to an emergency room: although the agency could not control who would need help or when, it could control its response. To do their jobs effectively, her staff needed accurate, real-time information to make informed decisions and to minimize case overloads that could impair judgment and lead to tragic outcomes.
Recently, the state of Illinois had launched a pilot program that used predictive analytics to identify children with a prior DCFS investigation who were at risk for serious injury or death, in the hope of reducing such disastrous occurrences.
However, Walker had begun to question the usefulness of the data-mining program. The new system seemed to obscure rather than improve sight lines, tying up limited resources while still enabling horrific acts of brutality that blindsided social workers and fueled public outrage. Walker wondered if a flawed program was better than no program at all. Could DCFS work with the analytics provider to reengineer algorithms and data sets to improve results, or should the agency discontinue the program entirely?