A hugely important work filled with knowledgeable insights, this book takes a hard look at the promises and pitfalls of big data. Mostly pitfalls. Written by an insider data scientist, the book's title riffs off the infamous Weapons of Mass Destruction (WMDs) of a decade ago, trading the Mass for Math. The clever theme continues with chapters labeled Bomb Parts (basics of mathematical modeling), Shell Shocked (the author's path toward recognizing the problem), Arms Race (going to college)...all the way through Collateral Damage and The Targeted Citizen.
Along the way author O'Neil examines how big data -
mathematical models - are now used to determine who gets into college,
how corporations target advertising to specific groups, whether you get
and keep a job, and assessing credit ratings and insurance risk. Through
insider interviews and personal experience, O'Neil documents how model
building often integrates the inherent biases of the people building the
models, as well as historical biases. In many different ways, and
through dozens of pertinent examples, it becomes clear that WMDs are
designed primarily to reduce costs and promote higher income for the
companies that use them.
Worse, WMDs reinforce societal
prejudices and stereotypes, targeting - even if sometimes
unintentionally - the poor and minorities, further driving them downward
and limiting opportunities for upward movement. The poor are kept poor
by reducing access to affordable loans, depressing credit scores, and
blocking job options through linkage to factors that are irrelevant or
biased. And because these models are black boxes both to the people held
back because of them and, often, the people administering and using
them, there often is no way to even know why rejections have occurred.
Without the model feedback seen in more useful models, these WMDs cause
their destruction with no hope of ever improving the algorithms used.
jumps from the financial crisis of 2008 to the removal of teachers
unfairly to how Google and Facebook influence behaviors simply through
their choice of what people see in their feeds - and who gets to see it.
models, algorithms, automation of processes, and online data collection
continue to become more prevalent, and potentially more destructive,
this book becomes essential reading. Its valuable insights, whether you
agree with everything the author suggests or not, are critical to our
informed discussion of what we want our future to look like.
Available on Amazon.