Required Reading (Especially for the NJDOE)
Cathy O'Neil's new book Weapons of Math Destruction: How Big Data Increases Inequality and Threatens Democracy is a must read for anyone living in our digital age; she's uniquely qualified to write this book, as she's an academic mathematician who earned her Ph.D from Harvard, worked for a hedge fund on Wall Street, analyzed big data for marketing start-ups and then became a political activist because she realized that a number of dangerous discriminatory algorithms are opaque, affect enormous numbers of people, and do unseen damage . . . she nicknames these WMDs . . . Weapons of Math Destruction, and she explains how these black box formulas evaluate creditworthiness, college rankings, our employability, our Facebook and Twitter feeds, and-- most significant to me-- teacher evaluations . . . and she spends a good portion of the book on just how irrational, absurd, and insanely unsound the models are that assess teacher performance-- the formulas might work if teachers taught ten thousand kids at a time, but for a class of 30 students, measuring how a kid did on a standardized test from one year to the next is essentially random (all the teachers know this, of course, even those of us who do not possess a math Phd. from Harvard, but it's nice to hear an expert explain the logic of why this is so) but apparently the NJDOE hasn't figured this out, and at the start of this school year, they increased the weight of standardized test scores in the evaluation model from 10% to 30% . . . so now, if a teacher works in a tested grade-- such as my wife-- one third of a teacher's numerical assessment is random . . . even if she teaches math and and can point out the many problems with the algorithm (a sociologist would cite Campbell's Law, of course, and also present a valid argument for why this change is absolutely inane) and I can't explain (without long strings of profanity) how incensed this makes me-- how utterly stupid the people at the NJDOE must all be, to enact this increase-- but I'm hoping that this book indicates a sea change in how we view these algorithms and formulas, and that people will learn enough math to understand how screwed up this is . . . and if the NJDOE changes the algorithm and writes a personal apology to me, confessing that they were totally ignorant of all math and logic, then I'm willing to forgive them, because even Bill Gates got it wrong with his charter school funding, he ignored the Law of Large Numbers and came to the conclusion that small schools were better than large schools, when the fact of the matter is that small schools have more statistical variance than large schools, because they have less students in them . . . so more of them will be better and more of them will be worse . . . but, of course, people may learn the truth and still not do anything about it-- we know that a later start time will improve test scores in high school, but the bus schedule prohibits this, and so kids show up at 7 AM, in a building without AC, ready to learn AP Physics . . . everyone knows this is not the best way to teach kids, but no one does anything about it, instead we purchase new software platforms so we can upload all the spurious data and crunch the numbers-- and there may be enough people in the NJDOE and other administrative capacities who love this idea so much, the idea that we're generating loads of numbers from standardized tests and evaluation algorithms, and they don't care that all the numbers are bullshit, because it's fun to have loads of "evidence" to evaluate and all this data perpetuates the idea that we need to pay people to look at it . . . anyway, I could go on and on, but read the book, it's revelatory . . . and if you don't feel like reading it, you can listen to her discussing it on Slate Money.