Wednesday, April 5, 2017

Child welfare’s 163-year addiction to bad science

Today’s child welfare system has its roots not in benevolence, but in bigotry.
Foster care began more than 150 years ago with a Protestant minister by the name of Charles Loring Brace. Brace founded an organization that still exists, New York City’s Children’s Aid Society.
Brace had seen the revolutions in Europe of 1848, and they terrified him. In particular, he was terrified of poor immigrant Catholics, whom he branded a “stupid, foreign criminal class” and the “scum and refuse of ill-formed civilizations.” He worried that “some demagogue might arouse their passions and fuse all the elements for a Parisian scene of riot and blood.”
And, of course, Brace explained, these Catholic immigrant parents were genetically inferior, passing bad “gemmules” on to their children. Fortunately, these ill effects could be reversed – by taking away the children and shipping them out to middle class Protestant farmers in the countryside.
Brace did just that. Between 1854 and 1929, his “orphan trains” took at least 100,000 such children. But the name notwithstanding, many of the children were not orphans. And many were taken without the knowledge or consent of their parents.
In 1990, I discussed some of this history in a letter to The New York Times. That prompted a touching and wonderfully ironic testament to the importance of family: a response defending Brace from his own great-grandson, Charles Loring Brace IV.
Bigotry? Heavens, no, Brace IV said. Those remarks about “scum and refuse” reflected merely a “somewhat genteel sense of superiority.” And you can’t blame great-grandpa for the whole “gemmules” idea – he got it from Charles Darwin!

The headline the Times put on this letter is: “Brace Took Up Bad Science in Good Faith.”

Bad Science in Arizona

Flash forward 163 years. The place is Arizona, scene of the nation’s longest sustained foster care panic. The sharp, sudden increase in children removed from their homes began in 2003. It might – might – have finally stopped in 2016. Arizona tears apart families at a rate 60 percent above the national average when rates of child poverty are factored in. Metropolitan Phoenix takes children at the highest rate among America’s big cities when poverty is considered.

And the state is remarkably creative about figuring out new ways to do it.
The latest plan: secretly tape interviews with those suspected of abusing and neglecting their children. Then run the recordings through a “Computer Voice Stress Analyzer” (CVSA) to see if the suspect is telling the truth.

Academics who have studied the technology … have dismissed it as ineffective. Kelly Damphousse, the dean of the University of Oklahoma’s College of Arts and Sciences, led a 2007 study for the Department of Justice. … “The software isn’t capable of distinguishing who is telling the truth,” Damphousse said …

“It’s a sham. … There is no scientific validity to the results,” said John J. Palmatier, a former police officer and criminal justice professor who has studied such technologies for 34 years. …  “They’re actually using it? I hope Arizona has deep pockets, because if the right attorney finds out, they’re going to pay big bucks,” Palmatier said.

In fact, the right attorney did find out. When family defender Gregg Woodnick blew the whistle, the Arizona child protective services agency decided it would not make clandestine use of CVSA after all. So that part of the plan never got off the ground. But the agency still uses it with the “permission” of the suspect.

The Ultimate in Bad Science

But of course, the classic example of child welfare’s never-ending love affair with bad science is “predictive analytics” in which computer software supposedly tells caseworkers which families are at high risk for abusing their children.

I have yet to encounter anyone who favors predictive analytics who is both an advocate of family preservation and does not run a child welfare system. But some reformers who do run such systems say, in effect:

§  Yes, we know that the Los Angeles experiment with predictive analytics produced a false positive rate of 95 percent.
§  Yes, we know about how predictive analytics reinforces racial bias in criminal justice, with a disturbing tendency to overestimate the dangerousness of the accused if they’re black and underestimate the danger if they’re white.
§  Yes, we know about the research in New Zealand that suggests similar problems in child welfare.

But, they say, we will avoid all these pitfalls. We will only use predictive analytics the right way.
I trust them to try their very best to do just that. But I don’t trust their successors, whoever they may be. And I don’t trust the successors to those successors. Because the typical child welfare agency is like the one in Arizona – chaotic, dysfunctional and ready to grasp at any straw that might get it off the hook the next time a child “known to the system” becomes front page news.

That’s why child welfare keeps taking up bad science in good faith.