Ben Smithgall

Welcome to the web blog

Software and data at Spotify. Previously with U.S. Digital Service and Code for America.

Home

Blog

Books

Projects

Weapons of Math Destruction

By Cathy O'Neil

Finished reading on December 4, 2018

O’Neil brings a needed practitioner’s voice into conversations critiquing the development and use of algorithms. Most of the critiques that I have been exposed to has been much more theoretical and nature and has not tackled the specifics so much. O’Neil spends a lot of time establishing her credentials: she worked as a quant at a large fund during the run-up to the 2008 crisis.

Perhaps the best chapter in the book comes right at the beginning as she documents her push towards disillusionment. I think that this is a valuable recollection of how industry can push people away. Besides O’Neil, only Zeynep Tefukci really seems to have a cogent set of theories about how things have gone off the rails with technology.

I do wish, however, that O’Neil had gone into a bit more depth with some of the “WMDs.” The book is clearly written for a general audience and thus sacrifices a bit of technical depth, but I think that was a missed opportunity. For those who need a primer as to how exactly technical and mathematical thinking can be misapplied to end up with nightmarish scenarios, this is a good start. However, I am still searching for a book that really captures in real detail the ways that these algorithms actually work in a detailed way.

It’s possible that we still don’t know because most of these companies operate their algorithms as complete black boxes: the only way to actually figure out whether or not things are bad is to just throw inputs at the box, examine the outputs, and guess as to what goes inside. But one thing O’Neil does talk about at length here is that you need to understand all of the parameters that are used in the models in order to understand how they are working.

★★★★☆