Computing Reviews
Today's Issue Hot Topics Search Browse Recommended My Account Log In
Review Help
Weapons of math destruction : how big data increases inequality and threatens democracy
O’Neil C., Crown Publishing Group, New York, NY, 2016. 272 pp. Type: Book (978-0-553418-81-1)
Date Reviewed: Nov 2 2016

It’s bad enough being afraid of clowns, but now it’s logistic regression that’s truly frightening.

Cathy O’Neil, who blogs as “mathbabe,” wishes that the details of algorithms that characterize people had to be public. She warns that today we forbid explicit discrimination but that lenders, landlords, and college admission officers can select people based on observations and models that may well have a discriminatory effect. She notes the apparent irrationality of much decision-making today: Why should your credit rating have anything to do with college admissions? Some of the algorithms seem obviously unfair and unethical; for example, is it really socially desirable that an election candidate’s campaign site show different promises to different readers based on an algorithm’s guess about the political leanings of the person accessing it? I understand the “efficiency” argument (you might want to know about economic policies, I might want to know about environmental policies), but it threatens the idea of an informed public, and it’s more difficult for journalists to spot contradictions than when a candidate says one thing to MSNBC and a different thing to Fox News.

What should we do about this? We can’t go back to the days when what mattered was who your father knew, and we wouldn’t want to. O’Neil argues that if the algorithms were public, they could be inspected and audited, and principles of fairness could be imposed by regulators. She points to successful examples such as your ability to correct mistakes in your credit rating. And this idea is in keeping with the “open data” movement, which calls for government data to be public so that everybody can look at the behavior of law enforcement, zoning decisions, and the like. The same “open data” idea also hopes to reduce mistakes in scientific results by requiring researchers to present their original data, as in the case of drug evaluations. We will need a larger number of journalists who understand statistical modeling, but teaching modeling would be a good thing generally since so much of our life now involves it; you can’t book a hotel room without some bot trying to decide how much you’re likely to pay for it. I do not look forward to the day when the intelligent gas pump talks to my car about how empty the tank is and decides it can charge me more per gallon because I’m about to run out of fuel.

This book is well written and extremely accessible; there aren’t any equations in it, nor does it go into the details of mathematical models. It’s a public policy argument, saying that fairness is a prime value of our society and has to be maintained against efforts to gain a little money by manipulating people. The author does point to places where algorithms and models are used to help disadvantaged people, and that’s fine with her, since they would be generally approved if they were public and discussed.

I can recommend this book to the general public, and those who do know what k-NN means should read this book as well to understand the implications of what they are doing.

More reviews about this item: Amazon, Goodreads

Reviewer:  Michael Lesk Review #: CR144896 (1702-0119)
Bookmark and Share
  Reviewer Selected
Editor Recommended
Organizational Impacts (K.4.3 )
Data Mining (H.2.8 ... )
Content Analysis And Indexing (H.3.1 )
Database Applications (H.2.8 )
Would you recommend this review?
Other reviews under "Organizational Impacts": Date
Enterprise networking
Grenier R., Metes G., Digital Press, Newton, MA, 1992. Type: Book (9781555580742)
Sep 1 1992
Managing computer impact: an international study of management and organizations
Bjo ., Eason K., Robey D., Ablex Publishing Corp., Norwood, NJ, 1986. Type: Book (9789780893913588)
Dec 1 1987
Computing and organizations: what we know and what we don’t know
Attewell P., Rule J. Communications of the ACM 27(12): 1184-1192, 1984. Type: Article
Jun 1 1985

E-Mail This Printer-Friendly
Send Your Comments
Contact Us
Reproduction in whole or in part without permission is prohibited.   Copyright 1999-2024 ThinkLoud®
Terms of Use
| Privacy Policy