I picked this up expecting a book about math and got a book about power that happens to be wearing math as a disguise, which is honestly a more interesting book than the one I thought I was getting. O’Neil, a former Wall Street quant who left finance disillusioned, walks through a set of algorithms that quietly run huge parts of American life, teacher evaluations, credit scoring, predictive policing, college rankings, and shows how each one launders bias through the appearance of objectivity.
her core definition of a WMD, a model that’s opaque, that scales, and that damages, is the taxonomy that stuck with me longest. it’s a genuinely useful lens, once you have it you start seeing it everywhere, the online ad targeting that quietly sorts people by wealth, the recidivism scoring that treats zip code as a proxy for danger. one reviewer said the book made them nod along at conversations they used to have about credit scores and car buying without ever questioning what was actually generating those numbers, and that’s exactly the effect it had on me too.
I do think the criticism I’ve seen that it stays light on the actual math is fair. this isn’t a deep technical dive, it’s closer to a tour of consequences, case study after case study of models doing harm, and if you already work in data science there might not be much new here for you. I’m not in that world, so it landed for me as genuinely eye opening rather than repetitive, but I understand why a specialist reader would want more under the hood.
the chapters on college rankings and predatory ads targeting people already in financial distress hit hardest for me personally, mostly because they’re systems I’ve interacted with directly without ever seeing the machinery behind them. O’Neil’s own quoted data scientist’s oath, the promise not to sacrifice reality for elegance without saying so, not to be overly impressed by mathematics, feels like it should be handed to a lot more people building these things.
four stars. more anecdote than technical manual, but the anecdotes add up to something that changed how I read every score attached to my name.
