My rating: 4 of 5 stars
I was in a used book store looking for one thing and found something else (isn’t that always the case?) I’d read Timandra Harkness’s Big Data and am ever interested in the subject, so of course I bought this. Right in the introduction, Ms. O’Neil hit my confirmation bias sweet spot: “Big Data has plenty of evangelists, but I’m not one of them.” If you are, maybe this will change your mind. Of course, that’s silly…if you are an evangelist, you are probably profiting from the Weapons of Math Destruction she talks about here and are resolute.
College rankings, teacher scoring, neighborhood profiling for crime prevention, evaluating how hireable someone is, loan applications, insurability, evaluating convictions for sentencing, credit scores, targeted advertising, …targeted election advertising…plus more…the message is the hammer to the forehead: Big Data and its exploiters are not your friend. Extensively researched, well composed, Ms. O’Neil’s scary narrative should be a wake up, but is lost in, well, the big data.
On the U.S. News & World Report‘s college rankings, Ms. O’Neil observes that the first report might have seemed sensible, but “as the rankings grew into a national standard, a vicious feedback loop materialized.” She continues: “The trouble was that the rankings were self-reinforcing.” This correlates with an observation of mine back in 1998 on a report with a couple of lines tracking retention of women and minorities at the senior ranks in a Navy group: if you make the tracking a point because you perceive there is a problem, there will always be that problem.
In a Certified Public Manager course, our class was asked a “what if” scenario if budgets had to be cut significantly and the law enforcement participants said the first thing to go would be investigating robberies and small crime, as manpower would be limited and the more serious ones would take priority. The opportunists have tried establishing a foothold already by developing software that profiles neighborhoods to target for “nuisance crimes” – ostensibly to better use resources for the “bad” areas instead of the “good” ones. Pre-Minority Report (which she does mention as a lead in).
Before the inane No Child Left Behind directive, the inept [my word] Reagan administration released a flawed report based on erroneous data interpretation that blamed teachers for big drops in SAT scores. Thanks to the Bush mistake, teacher evaluations became part of the criteria of defunding and some of those evaluations were “based almost entirely on approximations that were so weak they were essential random.” This is an example of a WMD that probably began with good intentions (like the college ranking) but the data weren’t understood well and the model was broken before it left the shop.
This book was published in 2016, so was written before the horror show of the 2016 US election and the subsequent revelations of deliberate hacking, collusion, targeting by states and stateless groups, so Ms. O’Neil’s analysis of targeted political advertisements is rather benign compared to the later all out efforts to sway and undermine the pseudo-democracy. She recalls Romney’s gaffe at not realizing some people were present to his candid remarks about 47 percent of the population being takers who were not part of the demographic he was talking to. She says that “very likely cost [him] any chance he had of winning the White House.” Fast forward four years to the facts-don’t-matter right wing supporters and I’m guessing she’s as dumbfounded that decency was thrown out the window, but not surprised at the skill of the Big Data manipulators in electing a meme (to borrow from a 4chan denizen quoted in another book.)
On a composition note, this book uses the end note format trend I detest: uncited in the text, listed by page number and the identifying sentence fragment, the excellent note sources are hidden until discovered. And because they are not cited in the text, it makes for a second pass through that annoys more than enhances – first time through it would be good to call attention to a source (a superscript does not detract from reading); seeing them at the end, I’m not inclined to go back and hunt for the passage the cite supports, no matter how intriguing that note is. “Gee, I wonder if that paragraph had a note. [Wastes time] No? Oh well, back to the next paragraph.” Can you tell I think it is a bane and should be abandoned for traditional citation?
Regardless, this is a good book that needs to be read more widely.