So the other day I binge read the relatively recent book The Data Detective by Tim Harford.
It's a pretty fun, breezy read and I definitely recommend it. It's like a slightly more socially aware and skeptical version of How Not To Be Wrong, which is also a book I really like.
I feel like while How Not To Be Wrong functions better as a pop-math book I'd actually say that The Data Detective, which has almost no overt math content, functions as a very good pop-epistemology book.
It covers a lot of ways we can fail to understand the data around us, the dangers of motivated reasoning, the problems of reproducibility, and what statistical data is---and is not---useful for and just overall how do you know from stats what you should believe.
There is just a lot of good material here on how to frame your thinking and remain open to alternative explanations rather than falling into the trap of seeing what you want to see in your data.
So, yes, overall this is a book I'd recommend pretty highly. It's a pretty simple, accessible read and it's also made me do a lot of thinking about how it'd be really great to take this kind of lucid approach and mesh it with a more mathematical course on pragmatically using statistics and coding to analyze real public datasets. I have a colleague who was already at least kind of interested in that and I think we could build something really cool this coming year.