Tuesday, April 11, 2017

Big Data’s latest brainstorm: Target poor people and make parents suspects if they ask for help

Remember the good old days when the one thing almost everyone interested in child welfare could agree on was “home visiting”? Assign a trained worker to help “at-risk” new parents, sometimes even before the child’s birth. Then follow up with regular visits to the home.
When such programs follow a particular model, the Nurse Family Partnership, this kind of help actually is helpful.
Those whose top priority really is taking away more children find this acceptable since it widens, rather than narrows, the net of intervention into families. Advocates of family preservation find it acceptable because it’s strictly voluntary for the families.
About the only people to object, initially, were right-wingers with their paranoid ideas that this was a sneaky backdoor way for “big gummint” to spy on families.
Just because you’re paranoid…

But then along came self-proclaimed liberal Elizabeth Bartholet, who proved the adage “just because you’re paranoid, doesn’t mean they’re not out to get you.” Based on the criteria she sets out in her book, “Nobody’s Children,” her agenda is so extreme that I estimate it would require taking at least one million children from their parents every year.
That would be even before implementing a linchpin of her strategy, which I discussed last month: a very different version of home visiting. The Bartholet version is the right-wing nightmare come true: It would be universal and mandatory. In other words, a government-mandated spy in every living room. Bartholet specifies that the visitors would be mandatory child abuse reporters, and that the purpose of those visits includes “surveillance.” Indeed, that seems to be their primary purpose.

Of course that’s not going to happen – not because it’s Orwellian, but because it would be so expensive. But now, it turns out, Bartholet’s vision also was hopelessly low-tech. In the age of the latest fad in child welfare, predictive analytics, it no longer matters what the spy in the living room actually sees. Just making use of a home visiting program ratchets up the level of suspicion.

At least that seems to be the latest plan from software vendor SAS, according to a recent story from predictive analytics evangelist and Bartholet disciple Daniel Heimpel, publisher of The Chronicle of Social Change.

SAS is the company which used past cases to test a secret, proprietary predictive analytics algorithm in Los Angeles. SAS proclaimed it a rousing success even though 95 percent of the time when the algorithm predicted severe harm to a child the algorithm was wrong.

Now SAS is developing a new approach in Florida. This one targets poor people. It compares birth records to three other databases: child welfare system involvement, public assistance and “mothers who had been involved in the state’s home visiting program.”

So listen up “at-risk” new mothers: In the world of predictive analytics, the fact that you reached out for help when you needed it and accepted assistance on how to be better parents isn’t a sign of strength – it’s a reason to consider you suspect, and make it more likely that your children will be taken away.
So the conclusion is obvious: Mothers will have to turn down the help in order to protect their children from the risk of having to face the horrors of foster care. Oh, wait, that probably won’t work. Big Data entrepreneurs would simply respond by finding a database listing mothers who refuse home visiting, and count that against them too.
The only way to escape Big Data is to hide the pregnancy, avoid prenatal care and give birth at home. Yes, child welfare has found one more way to endanger children in the name of protecting them.
A dilemma for predictive analytics defenders

SAS’ approach also should create a dilemma for some defenders of predictive analytics, such as Heimpel, who can’t see why anyone would object to using it for targeting which families should get help, in particular home visiting. SAS is standing that on its head. Their new approach winds up using home visiting to target investigations.
SAS’ evidence that its new approach works consists of pointing out that a lot of those it targeted had repeat reports of child abuse. But the reasoning is circular. Part of the rationale for predictive analytics is that decisions now are too subjective and prone to bias. Logically, then, you can’t turn around and cite those same subjective, biased decisions as proof your approach works.
Indeed, reliance on prior reports as proof of accuracy was a key flaw in evaluations of New Zealand’s experiment with predictive analytics.

And you’re certainly not going to make the system less biased by tracking only poor people. That only magnifies the existing “surveillance bias” that makes poor people targets because they’re more visible to government agencies.
But hey, the news isn’t bad for everyone. Middle class child abusers are in luck! If child welfare systems adopt this latest Big Data brainstorm, middle class child abusers are likely to be a lower priority for investigations – and likely to have lower risk scores when a caseworker shows up at the door.

As for poor people who need help with child-rearing, I guess they’ll just have to hire nannies.