Sen. Murray: Agreement Near on Predictive Analytics Pilot

Sen. Patty Murray, D-Wash. (Image: Daily Kos)

Sen. Patty Murray, D-Wash., said yesterday at a meeting of the Senate Health, Education, Labor and Pensions Committee she believes committee members are closing in on agreement on a bill that would  pilot a predictive analytics program to improve screening tools for the child welfare system.

The bill, named the Using Data to Help Protect Children and Families Act, would provide $10 million in grants for up to five state or local governments to pilot predictive analytics models to help determine the children most at risk of maltreatment. Sen. Todd Young, R-Ind., noted during the meeting that between four and eight children die every day in the U.S. due to abuse or neglect.

“I’m hopeful we can find a bipartisan agreement that provides a path forward for this new pilot, and the safeguards that are necessary for any new unproven program like this,” said Sen. Murray, ranking member of the committee.

“Used well, predictive analytics can help child welfare workers make better decisions about child safety and wellbeing, but if used poorly, they may perpetuate and amplify existing problems like biases based on race or socioeconomic status, or urban/rural inequities,” she said. “I believe we all want this bill to purposefully work to reduce bias in order to move us to better risk screening tools for child safety and wellbeing, and to keep this unproven pilot accountable to that goal.”

“We’re getting very close to an agreement, and… I believe we can get there if we have a shared goal for that program,” added Murray. Sen. Lamar Alexander, R-Tenn., chairman of the committee, said members will continue to work together on the legislation.

  1. Anonymous | - Reply
    This story indicates that Senator Murray has a misunderstanding of the possible benefits of predictive analytics which has little or no potential for predicting child maltreatment deaths or for improving safety assessments in child protection, as already demonstrated in Illinois. When used in child welfare screening units, predictive analytics can categorize screened-in CPS reports ( and only screened- in reports) into risk categories, e.g., high, medium and low, and, by doing so assist understaffed agencies in prioritizing reports for investigation. It remains uncertain whether predictive analytics will perform this function significantly better better than already existing risk assessment tools. However, the best most ethically defensible use of predictive analytics is to target a population of high risk families for voluntary prevention/ early intervention services to be offered as soon after a child's birth as possible. For predictive analytics to serve this purpose, the federal and state governments have to make prevention oriented services available to eligible families prior to a CPS report. If there are few, if any, services to offer high risk families, a targeting tool is of little value. It should be noted that there are other good ways of targeting high risk families for services prior to a CPS report, for example, by offering services to all parents of infants receiving federally funded substance abuse and mental health services, and to families of children who have a pattern of DV, and to families who live in neighborhoods or communities with a high rate of concentrated poverty. The advantages and disadvantages of these approaches to targeting families for voluntary prevention services requires research to sort out, and the ethical dimensions of these approaches need to be discussed in a civil manner. Dee Wilson
  2. Anonymous | - Reply
    A perfect tool of social engineering and thought police action

Leave a Reply