This week’s Ask a ScienceBlogger question from On High arrived while I was out of town (see also last week’s results), and I’ve held off answering because I had a huge stack of papers to grade. Of course, time for responding has almost run out, so I guess I ought to say something…

The question of the week is:

“If you could shake the public and make them understand one scientific idea, what would it be?”

Most of the other answers have come in already (and I’m too lazy to link them all), but they divide into two basic categories: answers dealing with the process of science, and answers dealing with the tools of science. There are excellent arguments for both sides, but I’m going to go for the “tools” category, and say “Statistics.”

(Expanded answer below the fold.)

By this I don’t mean things like regression analysis and t-tests and the Central Limit Theorem (“Everything’s a gaussian”), I just mean basic ideas of statistics like the difference between “mean” and “median” and some elementary probability analysis. I realize this is still a pretty broad category, but the ways in which some elementary knowledge of statistics would improve life in general are almost too numerous to count.

Study after study has shown that, as a species, we’re really, really bad about assessing probabilities. It’s well known that people grossly overreact to the threat of horrifying but highly improbable events, while ignoring much more likely dangers that feel less scary. One of my favorite examples was my grandmother’s belief that radon in the basement led to my grandfather’s lung cancer, despite the fact that he mostly went down there to smoke cigarettes.

But if there’s a wrong way to approach a statistical matter, you can count on the general public to find it. Studies showing tenuous links between things will be latched onto as graven-in-stone TRVTH, with no attempt to account for reasonable error bounds, while solid, responsible research will be brushed off as just so much fuzzy math, because, after all, you can prove anything with statistics. And the manipulations needed to deceive people are rarely complicated or subtle– strategic choice of mean vs. median values is usually about all it takes.

Huge swaths of public policy would quickly become much more sensible if only the voters had an elementary grasp of statistics.

Of course, I say this as someone who gave a “How to Lie With Statistics” talk for a class on the election of 2004 (there’s also grainy QuickTime video of a news story about the class, if you scroll up from that post). I also just (finally) got around to reading Freakonomics while waiting for my flight out of Knoxville on Saturday, so I’ve sort of got statistics on the brain at the moment.