<

  An open-blog for finding and sharing the best audiobooks Login (Register) / Logout



Weapons of Math Destruction: How Big Data Increases Inequality and Threatens Democracy

Weapons of Math Destruction: How Mediocre Books Get Mediocre Reviews and Threatens Readers



Author: Cathy O'Neil
Narrator: Cathy O'Neil
Publisher: Random House
Length: 11 h 20 m
Genre: Self-Help & Popular Psychology
Published: 2014
Reviewer: Anonymous

Book Rating 
 

Audio Rating 
 

Overall
 

The author, Cathy O'Neil, does not seem to understand humans very well. I believe this book is about her journey through first trying to understand them through statistics to realizing that that simply is not possible. Superficially, Ms. O'Neil is writing about the poor power of prediction of mathematical models when applied to people, or how these models can amplify unconscious bias against certain minority groups. But what I think the author is really writing about is what a difficult time she has had understanding the people in her life. Like most autism spectrum researchers turned authors, we kick off with a great story about how "Big Data", if used improperly can have devastating effects. This is because (writes O'Neil) big data analysis is an automatic process once the model is built and accepted, that is, it is hard to know why a model "spits-out" a particular answer given a particular input, so mostly, it just needs to be either accepted or not. Unintended bias crops up when, say, a bank is trying to decide how much interest to charge a customer. Lets say our Bankers are honest and good and do not want to discriminate based on race so they strive to not include that type information in their model. Later they may find that (statistically) they are discriminating based on race, even if their model only includes variables like income, age, location, home size, past loans, etc. Thus, in a case like this, "Big Data" fails to produce an unbiased result, one which may never be caught unless another analysis is done after the fact, though this is rarely done in industry says, O'Niel.
I would not recommend this book. It is for no one. Not for data scientists, O'Neil's insights are not profound, they are interesting, but nothing new. Not for laymen, even given the excellent writing and (for the uninitiated) eye opening ideas, I don't think there are many in that category who care to know how these algorithms work but also don't already know that modern frequentist statistics is seriously flawed.
After finishing the book I right away watched her TED" talk and was disappointed to find it almost word for word the opening book chapter, same with a few other talks I tried. It was a good book, try it if you want, but to be honest, the last two chapter I did not want to finish and were a bit of a chore.