Weapons of Math Destruction

 Weapons of Math Destruction: How Big Data Increases Inequality and Threatens Democracy, Cathy O'Neil, Crown Publishing, 2016

Big data and algorithms are supposed to be the "savior" of our modern world. With them, a corporation, or a government, is supposed to be able to measure and analyze nearly anything. What if those algorithms are very flawed?

Among the suggestions to fix American education is to get rid of bad teachers. Standardized test scores are one way to find those bad teachers. What if the students didn't learn the basics of math, for instance, in a lower grade? What if the teachers in that lower grade blatantly corrected the standardized tests, before submitting them, to make themselves look better? If the test scores for a class are not as good as the algorithm predicted, then that teacher is out the door.

Crime prediction software sounds like a godsend to cash-strapped police departments. Why not concentrate resources in areas where there is predicted to be a better chance of crime, instead of everywhere? If the police department includes "nuisance" crime, like underage drinking or pot smoking in public, the algorithm will send units to that neighborhood on an increased basis. If it happens to be a minority neighborhood, and is otherwise law-abiding, the residents can expect more incidences of "stop and frisk." Again, changing that algorithm is not possible. 

At work, it is not possible to change the algorithm that makes the schedule for the employees because this person has transportation issues or that person has child care issues. "Clopening" is when an employee of Starbucks, for instance, closes the store at 11 PM, then has to return in a few hours to open up the next morning, and work a full shift. 

Algorithms have their good and bad parts. The biggest bad part is that there is no way to change them, and get them to conform to the real world. Written by a data scientist, this book is a big eye-opener and is very much worth reading.