I’ve followed Tim Harford’s blog for many years. He’s an engaging and interesting writer, and exudes a curiosity about the world around him. I was expecting to enjoy his latest book The Data Detective, and I was not disappointed even though I’d read many of the anecdotes he shares. The book is subtitled: “Ten Easy Rules to Make Sense of Statistics”. You might think that hard data and statistics are in opposition to anecdotal stories, but the blend in Harford’s writing illustrates why both are important.
The rules are really rules-of-thumb. That’s important, because things are often more complicated and more interesting – something that an iron-clad hard-and-fast rule cannot capture. I could wax unpoetically about this in theory, but instead I will pick out four rules that caught my eye along with tidbits of personal commentary. For Harford’s excellent writing, I recommend reading his book in full.
Rule One. Search Your Feelings. The chapter begins with a quote from Darth Vader: “Search your feelings, you know it to be true.” The opening anecdote is about an art dealer, a forger, and a top Nazi commandant. I won’t spoil the story. The key point that Harford makes in this chapter is that stood out to me: “Experts are not immune to motivated reasoning. Under some circumstances their expertise can even become a disadvantage.” All this made me think about times I had gotten hot-around-the-collar about some topic. This doesn’t happen often, making these moments all the more memorable. In a bad way. Because it made me harden my position while spewing out information to support my digging in, and less open to listening. It clouded my judgment. In hindsight, I’ve been wrong some of the time. And even when I wasn’t wrong, I was less right than I thought.
Rule Three. Avoid Premature Innumeration. The opening quote is from Deep Thought via Hitchhiker’s Guide to the Galaxy: “Once you do know what the question actually is, you’ll know what the answer means.” I was reminded of the pitches and arguments I’ve made throughout my career as a faculty member and administrator. They’ve often “data-driven” because I try to take full advantage of the scientific veneer, justified or not. More often than not, I succeed in getting what I want. Sometimes I don’t, but I feel smug about my parting Cassandra-like warnings of the impending doom. Needless to say, hindsight reflection on the way I’ve marshaled statistics isn’t as flattering. I haven’t told lies or damned lies that I’m aware of, but I had blinkers. I was also reminded that my forecasting ability isn’t as good as I think it is.
Rule Seven. Demand Transparency When the Computer Says No. I won’t tell you the opening quote except that it comes from HAL9000 from 2001: A Space Odyssey. You can guess what this is about. I’m a computational chemist who thinks about problems in complex systems; I submitted a grant on machine-learning approaches a few months ago (hope it gets funded!); I’ve been reading about algorithms and their limitations. The anecdotes in this chapter are particularly interesting – from the errors in determining “normal” body temperature to a story about a risk-assessment algorithm aimed at predicting re-arrest rates of offenders. I’m suspicious whenever anyone tells me the inner workings of their analysis are a little complicated to explain. Try me! But then I do the same thing when I’m presenting, sometimes because I’m trying not to get sucked down into a rabbit black hole, sometimes because I’m sensitive to time-constraints, sometimes because I’m obfuscating on purpose, sometimes because I don’t understand things as well as I should. Now that’s complicated. I’ll quote Harford: “Trust should be discriminating: ideally we should trust the trustworthy, and distrust the incompetent or malign.” Easier said than done though.
Rule Ten. Keep an Open Mind. The chapter opens with a comparison between two economists who lived through the Great Depression in the early twentieth century: Irving Fisher and John Maynard Keynes. Most of us have heard of the latter, but not the former. There’s a reason for that (read Harford’s book for the details). My reflection on this story is that if you’re successful early on, and have “often been right”, you’re ripe to colossal close-minded failure. A few setbacks along the way provide a dose of reality that can hopefully make one more circumspect. It was useful to look back at my career, recognize some of the main ups and downs, and learn some open-mindedness. (There’s also an interesting story about the Millikan Oil Drop experiment and the process of determining a fundamental constant, which reinforces the point of how we humans filter data to our own liking.)
Harford’s epilogue is titled: “Be Curious”. He sketches an outline of how we might reduce political polarization. I mention this to pique your curiosity. But what I’ll reflect on is the distinction Harford makes between scientific curiosity and scientific literacy. I’ve been involved in two major core curriculum designs – one from scratch for a new college, and one major redesign at my home institution. There was much wrangling over how we educate college student to be scientifically literate citizens; there wasn’t as much talk about how we encourage scientifically curious citizens. I think we want both, and I think they go hand-in-hand. I also think that many of the contemporary science education discussions muddle the distinction. This is something I’d like to chew on a bit more, before possibly blogging about it at a later time.
All in all, I highly recommend The Data Detective!
For a selection of previous blogs on books about statistics, see:
No comments:
Post a Comment