Sunday, January 28, 2018

Predictive analytics in child welfare: Sanitized for libertarian consumption

A writer who was barred from blogging for an education journal after writing a column widely viewed as racist now assures us that there is no racism in child welfare – and predictive analytics will correct the racism that isn’t there.

Don’t you feel better already?

An odious piece of child welfare legislation known as the Adoption and Safe Families Act enshrined a “take-the-child-and-run” mindset in federal law leading to an increase in the number of children torn from their homes. It also increased the number of children who “age out” of foster care with no home at all. The former was intentional, the latter was merely predictable. 

A few years after ASFA became law one of those who claims responsibility for writing it, Richard Gelles,  couldn’t resist a little gloating. Gelles is the former dean of the School of Social Policy and Practice at the University of Pennsylvania. In an interview in 2000, he said:
 Initially, this was just supposed to be a safe families bill, not really an adoption bill at all. The adoption component was a way of sanitizing the bill, to make it more appealing to a broader group of people. Adoption is a very popular concept in the country right now. [Emphasis added.]

Of all those who advocate for a take-the-child-and-run approach to child welfare, Gelles is probably the second most extreme – I’d give the edge for #1 to Elizabeth Bartholet, but it’s close. (For a brief critique of Gelles, see this commentary from his faculty colleague, (and NCCPR board member) Prof. Dorothy Roberts.)

So it is appropriate that when right-wing writer Naomi Schaefer Riley needed a way to sanitize the use of predictive analytics in child welfare for a crowd that might be suspicious – libertarians – she turned to Gelles. Gelles was her primary source in an article touting predictive analytics she wrote for the libertarian magazine Reason.

Predictive analytics uses computer algorithms to tell government authorities things like which alleged criminals should get bail, who to investigate as an alleged child abuser and, if some of its strongest backers get their way, which children should be taken from their parents and consigned to foster care.

The failure of predictive analytics in child welfare is discussed brilliantly in this article from Wired, (an excerpt from the new book Automating Inequality), so I won’t discuss that failure in detail again here.  Rather, I want to focus on how Riley tries to win over libertarians.

● She sandpapers down the rough edges of her own take-the-child-and-run rhetoric, which can be found in its raw form in columns for Rupert Murdoch’s New York Post. For Reason, she finds others to say the words with a bit more finesse – though in at least one case, discussed blow, the result is laughable.

● She makes a little room for those of us who dissent, quoting us in what opinion writers commonly refer to as “to be sure” paragraphs (as in “To be sure, the side I don’t agree with says … but of course they’re wrong because …”)

● She – and Gelles – even try to sell us on the notion that they care about errors in all directions  and they really, truly want child welfare systems to use predictive analytics because supposedly it will curb all kinds of error.  So Riley writes:

Every day some kids are forcibly taken from their parents for the wrong reasons while others are left to suffer despite copious warning signs.

Strangely, however, I have been unable to find a Riley column in the Post about the former, but there is plenty about the latter. Indeed, she is a master of the standard technique of the take-the-child-and-run crowd – take horror stories and extrapolate.

Thus she writes that David Hansell, the current commissioner of New York City’s child welfare agency, the Administration for Children’s Services, “has to clean up the mess left by his predecessor …” based on the fact that on that predecessor’s watch children “known to the system” died.  That also happened on the watch of every other ACS commissioner - and every leader of every large child welfare system. In fact, under the leadership of that predecessor Riley does not bother to name – it was Gladys Carrion - child safety in New York City improved. It’s gotten worse, (or should I say messier?) under Hansell.

Riley on race

Most disturbing, given her desire to reassure us that there is no racial bias problem with the use of
predictive analytics, is Riley’s general take on racial issues.

Riley was barred from blogging for the Chronicle of Higher Education after writing a column in 2012 calling for eliminating the entire field of Black Studies – because she deemed three doctoral dissertations “left-wing victimization claptrap.” As others pointed out, that conclusion, is based only on the titles and summaries run in a Chronicle news story.    

With that as the basis, Riley wrote:

… the entirety of black studies today seems to rest on the premise that nothing much has changed in this country in the past half century when it comes to race. Shhhh. Don’t tell them about the black president! 
… If these young scholars are the future of the discipline, I think they can just as well leave their calendars at 1963 and let some legitimate scholars find solutions to the problems of blacks in America. Solutions that don’t begin and end with blame the white man.

Of course Riley wrote that in 2012. She couldn’t have known who would be the next president, and about the racial hatred he would unleash and continues to encourage.  But in 2017, she was still singing from the racism-is-so-over hymnbook.  In a column called “How liberals are killing the NAACP” she writes:
 African-Americans have full legal rights. Hate crimes are anomalies. Black people are running corporations, universities and until recently the White House.

So it should come as no surprise that Riley is a member of what I call child welfare’s “caucus of denial” – the group that believes child welfare is magically exempt from the racial bias that permeates every other aspect of American life. But unlike the liberals in this caucus, who tend to limit the denial to child welfare, Riley goes further, writing:

Just as we cite statistics about incarceration and police intervention without actually asking who is committing crimes, so we blame ACS for getting involved in the lives of too many black and Hispanic children without asking why that might be.

Actually, that question has been asked and answered in study after study.  Some conservatives acknowledge as much. It seems Riley just doesn’t like the answers.

Riley’s “denial” comes in a column trashing the New York Times story about foster care as the new “Jane Crow” – a story that also explains “why that might be,” citing exactly the sorts of cases of needless removal which, in her Reason article, Riley claims to be concerned about.  In the Post Riley sneers at the Times story in her lead:

Everyone’s a little bit racist, even the employees of the Administration for Children’s Services. At least that’s what The New York Times would like us to believe.

In her Reason article, Riley offers up the same message about the Times story, without slathering it in snide. Instead, she quotes Bill Baccaglini, president of  The New York Foundling one of the private child welfare agencies paid by the city to provide foster care.

Some of my best friends are …

First, Riley and Baccaglini fall back on the 21st Century equivalent of the 1960s “some of my best friends are …” line: The system can’t be racist (at least not in New York City), they say, because most of the caseworkers are nonwhite. This ignores the way institutional racism works and it ignores intra-ethnic prejudice, the sad fact that within almost every racial and ethnic group there are biases.

Anthropologist Tina Lee, who wrote an outstanding book about the New York City child welfare system, addressed this during a recent podcast. Said Lee:

In New York City caseworkers often are also women of color; they may come from a slightly different class background. So one of the lessons of this is that these racial biases are not overtly about white people being overtly prejudiced and seeking to control these families. It’s more that we just have these assumptions that are often unconscious. [These are] unconscious biases in our general culture that you don’t escape even if you are a woman of color working as a caseworker.

And then it gets really weird. Baccaglini cites the following as evidence that there is no racial bias in New York City child welfare:

You couldn’t even consider race a variable. It’s a constant.  All of the kids who come into this system, unfortunately, are nonwhite.

Wait. You mean this is how we’re supposed to know the system is NOT racist?

Baccaglini then offers up the standard party line of the denial caucus:

The racial discriminatory aspects of the system happened well before [child protective services involvement] with our opportunity structure … The fact that the mom in the South Bronx cannot get decent medical care; the fact that the mom in the South Bronx cannot get a good job; the fact that the mom was put into an [individualized educational program] and never got a degree and then had a child.

In other words, Baccaglini is telling us, there used to be racism in America and that made Black people bad parents!  (The real story of how child welfare treats families in the South Bronx can be found here.)

The idea that the solution here might be helping the mom with job training, health care and education apparently is not on the table.

A poor understanding of poverty

And that leads to another problem with Riley’s attempt to lure libertarians into a love of big government: misstating the relationship between poverty and the child welfare system.

Adopting a voice of reason tone for Reason that differs from her Post persona, Riley tells us:

There is a bit of a chicken-and-egg problem here. Poverty is highly correlated with abuse. There are a variety of reasons for that, which can be difficult to untangle. Poverty causes stress in marriages and other relationships, and sometimes that stress is taken out on kids.

But that’s only a small part of the story. Yes, of course poverty exacerbates stress and that can lead to more actual abuse. But Riley ignores the larger issue, the confusion of poverty itself with neglect, thanks to broad, vague laws that often define neglect as lack of adequate food clothing or shelter.

Once again, she is more direct in the Post, where she can appeal directly to her right-wing base. There, she offers a critique of social work education similar to her assessment of Black Studies – but this time she relies on course catalogues instead of dissertation summaries.

She finds it deeply disturbing that one school of social work offers a course exploring “poverty in the context of oppression, diversity and social justice” and another has one that helps students “challenge bias, prejudice and forms of discrimination…” This supposedly proves that social work education often is “boxed in by political correctness in an industry where that can be deadly.”

Given all this, it’s no surprise that in Reason she echoes the argument of the white middle-class foster parent who extolled predictive analytics in child welfare in an article for The New York Times Magazine: There is no racial bias in child welfare – and predictive analytics will fix the bias that already doesn’t exist!

Whitewashing failure in Los Angeles

Like the Times Magazine writer, Riley also has a problem dealing with the failure of the first predictive analytics experiment in Los Angeles, known as AURA. The Times Magazine story solved the problem by ignoring it altogether.  Riley takes a more confusing approach. She touts its alleged success, but then notes that

Nonetheless, [the Los Angeles child welfare agency] ultimately concluded that AURA was "fatally flawed."
Riley’s only explanation:

Perhaps because they realize how sensitive the program is, and how much relying on the wrong model could have undermined public confidence, they are self-reflective and critical in a way you might not expect from government bureaucrats.

Yes. That must be it.  It can’t possibly be because the experiment had a false positive rate of 95 percent – that is, 95 percent of the time, when the model predicted something terrible would happen to a child, it didn’t.  Riley never mentions this in her story.

So add Naomi Schaefer Riley to the long list of take-the-child-and-run advocates who are pushing predictive analytics, such as Gelles, and Bartholet. 

And almost all of them now are trying to persuade us that, really, truly, predictive analytics will curb the needless intrusion into the lives of Americans and the wrongful removal of children, something none of them has shown any indication of caring much about before.  It reminds me of how congressional Republicans suddenly embraced the Children’s Health Insurance Program, when they thought it would help them score political points during the debate over the recent government shutdown.

There is nothing libertarian about predictive analytics in child welfare. It’s just big government conservatism.