For those interested in knowing more about the problems with predictive analytics in child welfare, NCCPR’s full analysis is available in our report Big Data Is Watching You.
I will leave it to others to try to guess what the election of
Donald Trump means for child welfare policy, aside from pointing out that in
addition to all the other reasons to worry, as far as I know the only one of
his close advisers who’s ever thought about the topic – Newt Gingrich – has
suggested throwing poor people’s children into orphanages.
But we
shouldn’t let the postmortems go by without child welfare learning from one of
the biggest losers on election night: predictive analytics.
This was the year when an
amazing number of organizations, from FiveThirtyEight to The New York Times, to the Princeton Election
Consortium were all crunching numbers, assessing data points and putting it all
together into algorithms that were going to tell us who was going to win the
presidential election. The overwhelming consensus: Hillary Clinton.
The
same predictive analytics number crunchers also had algorithms to tell us which
party would wind up in control of the Senate. The not-quite-as-overwhelming
consensus: the Democrats.
Speaking
just for myself, I wish all those predictions had been correct. Instead, we are
left with two YUGE “false positives.”
“American voters just
tossed an ice cold bucket of reality on those who argue that Big Data is here
and now, and ready to run everything,” writes Forbes columnist John Carpenter.
It was a rough night for number
crunchers. And for the faith that people in every field … have
increasingly placed in the power of data. [Emphasis added]
If all
those number crunchers can’t figure out one presidential election, why should
anyone trust predictive analytics to tell us which parents are going to harm
their children? Yet that, of course, is what some in child welfare seriously
want us to do.
The
hubris behind that effort is astounding, and dangerous. The various data gurus
who got the election forecasts wrong suffer nothing worse than public
humiliation. Wrongly predict that a parent is going to abuse a child and
there’s an excellent chance the child will suffer the enormous trauma of
unnecessary foster care – and workers will be overloaded with all those
needless removals, leaving them less time to find children in real danger.
The predictive analytics problem goes beyond elections
Of
course, one could argue even a slew of bad election predictions (it wasn’t just
one election, Big Data was wrong about several states) is not, alone, enough to
say we should not press forward with letting algorithms tell us when to tear
apart a family. The botched election predictions could be thought of as just a
horror story – and we all know people in child welfare never, ever base policy
decisions on horror stories, right?
But the Times says the lessons go much deeper:
… undercutting the belief that
analyzing reams of data can accurately predict events. Voters demonstrated how
much predictive analytics, and election forecasting in particular, remains a
young science …
data science is a
technology advance with trade-offs. It can see things as never before, but also
can be a blunt instrument, missing context and nuance … But only occasionally —
as with Tuesday’s election results — do consumers get a glimpse of how these
formulas work and the extent to which they can go wrong …
The danger, data experts say,
lies in trusting the data analysis too much without grasping its limitations
and the potentially flawed assumptions of the people who build predictive
models.
Flawed assumptions, built
into the models, were the root of the rampant racial bias and epidemic of false
positives that an in-depth study done by ProPublica found
when analytics is used in criminal justice. Prof. Emily Keddell found much the same when
she examined bias and false positives specific to predictive analytics in child
welfare.
The Times story also includes a lesson for those who
insist they can control how analytics are used – those who say they’ll
only use it to target prevention – not to decide when to tear apart families:
Two years ago, the Samaritans,
a suicide-prevention group in Britain, developed a free app to notify people
whenever someone they followed on Twitter posted potentially suicidal phrases
like “hate myself” or “tired of being alone.” The group quickly removed the app
after complaints from people who warned that it could be misused to harass
users at their most vulnerable moments.
The
failure of predictive analytics in this election should be one more warning not
to data-nuke poor families. It should, but, as usual when the arrogance of
software companies meets the arrogance of the child welfare field, it probably
won’t.
