A second post on Bergstrom and West’s enjoyable book, Calling Bullshit. They tackle the problems we face with the amount of nonsense around in the information age. The book contains much useful advice on spotting problems in black boxes and avoiding zombie statistics.
You Can Spot Nonsense Even In Black Boxes
I liked their advice that “you usually don’t have to open the analytics black box in order to call bullshit on the claims that come out of it” (Bergstrom and West, 2020, page 42). This is important because in many areas where quantitative models are used it is easy to not understand all the details immediately. Yet, the challenge is that you shouldn’t dismiss models because you don’t understand them perfectly. There are plenty of examples of great things most of us don’t understand but work. For example, I don’t know much about medicine but I don’t assume it is all useless because of this.
That said, sometimes the claims from models really are pretty silly. How to know the difference without doing the math? Often it is enough to look at the inputs and outputs. Bergstrom and West talk a lot about deficient training data. If you use bad data to train a model then what it predicts won’t be very good. Similarly think hard about the claims made in respect of model outputs, something they simply aren’t plausible with what we know of the world.
Incompetence Is Often the Reason
I also enjoyed the sensible advice to, “Never assume malice or mendacity when incompetence is a sufficient explanation” (Bergstrom and West, 2020, page 136). A lot of BS is out there because those spewing it don’t know any better.
Zombie Statistics
Zombie statistics is a great notion. It is pretty fascinating because you often get them in marketing. Information is provided which has no obvious provenance. Who did the original research to find the stat? No one knows. Maybe there was a glimmer of truth in there initially but now no one can say where the ‘fact’ comes from. Don’t be fooled by a precise number into thinking that it is a meaningful number. “Without knowing the source and context, a particular number is worth little” (Bergstrom and West, 2020, page 100).
For more on this book see here.
Read: Carl T. Bergstrom, and Jevin D. West, Calling Bullshit: the art of skepticism in a data-driven world. Random House, 2020.