Ellenberg is a mathematician working (I think) in number theory. This book doesn’t deal with that so much. It’s more along the lines of “Innumeracy”. That is, we get a lot about how to (as the title suggests) use mathematics and statistics not to be wrong. Inevitably along this path we get some information on how to spot when they are being used incorrectly or deceptively as well. And there is a constant undercurrent of, yes, mathematics is useful to the real world, along with a recognition that mathematics as taught to the masses does not really do a good job of revealing this.

What sorts of things are covered, then? One of my favorite pet peeves - not everything is a line - is hit straight away. There is a lot of discussion of statistical reasoning, how it is used and misused, and what some of the trouble spots are. Of course we see probability and the interplay between things being unlikely yet there being so many things that some unlikely things are likely to happen (and then get too much importance attached to them). I believe “Innumeracy” had an example concerning the likelihood that you actually have a disease given that you test positive for it (lower than you think) - this has a corresponding example about Facebook and terrorists. Both examples are there to show how easy it is to get conditional probability wrong.

There is also, as should be the case in just about all of this type of book these days, discussion of voting mathematics. Although it stops short of Arrow’s theorem, it gets most of the way there, and there are concrete examples of not only how traditional voting can produce results that seem wrong, but also how variations (such as ranked choice) can do the same. Arrow’s theorem basically says that you have to pick your poison here, and while I’d prefer the poison of ranked choice between the two. And while I like other ideas even more, it’s important to know that whatever one is picking it is mathematically proveable not to be perfect (i.e. such does not exist).

Ellenberg ends with an argument against the view that to look at things in the ways he described, which often include recognizing and embracing inherent uncertainty, as a sort of namby-pamby, wishy-washy take on the world. I agree - the idea in my view is to take in truth as closely as we can, and that to me means dealing with the fact of uncertainty. An example that Ellenberg does not use to make it concrete - there are statistical systems for predicting the outcomes of (say) basketball games. We can collapse the information that they give to just judge them on how often they predict the winner of a game. However, that is a bad way to judge. Better would be to take a collection of games that they say are (for example) 70-30 shots. Does the favored team win about 70% of the time? Do the same for other brackets. Some of them may even provide more detailed predictions. How close do they come to those? That’s a fairly trivial example, but gets at the heart of how we should really evaluate models in a stochastic world.