Monday, November 28, 2016

When the good guys ride into the sunset, leaving their algorithms behind

The push to embrace “predictive analytics,” an unreliable, raciallybiased method in which algorithms “inform” the decisions of caseworkers – decisions such as when to tear apart a family – is led largely by the same people who deny the very existence of a racial bias problem in child welfare, want to see far more families torn apart, or both.
They seem to be willfully blind to the enormous inherent risks of unnecessary foster care placement, including the risk of abuse or evendeath in foster care.
But there are a few reform-minded leaders in child welfare who want it as well, for themselves. These reformers argue in effect: “You can trust us. We’ll only use our vast powers for good.”
I do trust them. But I don’t trust their successors.
That brings me to another lesson from the 2016 election, a lesson that extends beyond predictive analytics to a wide range of child welfare issues.
The lesson comes from how President Obama handled key national security issues. He believed the George W. Bush administration’s ends-justify-the-means approach to the “war on terror” had gone too far. But instead of pushing to change laws to make it harder for any president to abuse these powers, he opted for an approach that amounted to: Trust me.

Limits of the “Have-It-Both-Ways” Approach

The New York Times aptly described it as “President Obama’s have-it-both-ways approach to curbing what he saw as overreaching in the war on terrorism.” According to the Times:

Over and over, Mr. Obama has imposed limits on his use of such powers but has not closed the door on them — a flexible approach premised on the idea that he and his successors could be trusted to use them prudently. [Donald] Trump can now sweep away those limits and open the throttle on policies that Mr. Obama endorsed as lawful and legitimate for sparing use, like targeted killings in drone strikes and the use of indefinite detention and military tribunals for terrorism suspects.

 And even in areas where Mr. Obama tried to terminate policies from the George W. Bush era — like torture and the detention of Americans and other people arrested on domestic soil as “enemy combatants” — his administration fought in court to prevent any ruling that the defunct practices had been illegal.

I don’t know the average tenure of people who run public child welfare agencies, but I doubt it’s eight years, or even four. The reformers clamoring to get their hands on predictive analytics algorithms may intend to use them with great restraint. But there is nothing to stop their successors from using analytics any way they want.

Lessons Beyond Predictive Analytics

The lesson should extend beyond predictive analytics. Child welfare systems are unique in state and local government in the extent to which they wield vast power with little accountability.
§  Child protective services workers can search homes – and strip search children – based on no more than an anonymous tip.
§  They can walk out with the children entirely on their own authority, or ask law enforcement to do it for them, without any hearing beforehand.
§  When there finally is a hearing, the indigent parent – and most are indigent – may not get a lawyer at all. If she does, it’s likely to be an overloaded public defender she just met in the hallway five minutes before the hearing.
§  The standard of proof to hold the child indefinitely is not “beyond a reasonable doubt” as in a criminal case. The standard usually is merely “preponderance of the evidence” – slightly more likely than not – the same standard used to decide which insurance company pays for a fender bender.

The few reform-minded leaders of child welfare systems know all this.  And they know how this vast power, when abused, harms the children needlessly taken, and how it overloads workers so they’re less likely to find children in real danger. But their answer often is the same as President Obama’s approach to that other area with vast power and little accountability, national security: self-restraint.
Twenty years ago I met one such reformer. She’d spent years curbing needless removal in her state. But shortly after she left, when a new governor took office, there was a high-profile fatality. The governor panicked, and effectively ordered his new child welfare agency chief to do the same. “I was amazed at how quickly it could all be washed away,” she said.
So reformers who are serious about a lasting legacy need to get serious about curbing their own power. They should be urging lawmakers to, among other things, raise the standard of proof in child welfare cases and curb the ability of workers to misuse their “emergency” authority to remove children. (NCCPR’s full DueProcess Agenda is available here.)

Most important, they should be creating institutional providers of high quality defense for families, along the lines of what’s been done in New York CityThis is not a strategy to get “bad parents” off; it helps craft alternatives to the cookie-cutter service plans typically imposed on families even by relatively good child welfare agencies.

The best way to create a legacy of reform is to give families themselves the power to fight for it.