The Clear-eyed View

I've been meaning to blog about Blink, Malcolm Gladwell's new book. In the author's words:


When you meet someone for the first time, or walk into a house you are thinking of buying, or read the first few sentences of a book, your mind takes about two seconds to jump to a series of conclusions. Well, "Blink" is a book about those two seconds, because I think those instant conclusions that we reach are really powerful and really important and, occasionally, really good.

You could also say that it's a book about intuition, except that I don't like that word. In fact it never appears in "Blink." Intuition strikes me as a concept we use to describe emotional reactions, gut feelings--thoughts and impressions that don't' seem entirely rational. But I think that what goes on in that first two seconds is perfectly rational. It's thinking--its just thinking that moves a little faster and operates a little more mysteriously than the kind of deliberate, conscious decision-making that we usually associate with "thinking." In "Blink" I'm trying to understand those two seconds. What is going on in inside our heads when we engage in rapid cognition? When are snap judgments good and when are they not? What kinds of things can we do to make our powers of rapid cognition better?

Gladwell cites study after pop-culture study to answer these questions. It's fascinating reading about taste-tests, gamblers' sweat, the Implicit Association Test, and anecdotes from various experts (art, food, marriage, and more) about their honed first impressions.

For instance, Gladwell recounts Gawande's story of the experienced firefighter who "sensed" a floor was about to collapsed (previously blogged about last March -- his intuition was actually guided by subtle clues from the fire that he processed subconsciously).

Then there's the marriage researcher who's watched so many videos of couples, and studied outcomes for so long, that he can process tiny details of intonation and body language and predict which couples will last, after just a few seconds of watching.

One study of interest to doctors: a group of psychologists were able to predict which surgeons would be sued for malpractice, not by looking at hospital citations, or training sites, or previous suits. Rather, they make predictions based on snippets of patient-doctor dialogue, with the voices stripped of high-pitched frequencies (so as to make the actual content unintelligible). In other words:

The judges knew nothing about the skill levels of the surgeons... They didn't even know what the doctors were saying to their patients. All they were using for prediction was their analysis of the surgeon's tone of voice... Malpractice sounds like one of those infinitely complicated and multidimensional problems. But in the end it comes down to a matter of respect, and the simplest way respect is communicated is through tone of voice... (page 43)

The trouble is, my first impression of this study, blogged about in May, was that it was bunk. Shazaam agreed! Maybe I'm not enough of an expert to trust my formative impressions, but there still seem to be many potential confounders -- is a dominant tone from the surgeon related to a patient's belligerent body language? Or to the degree of bad outcome? The paper (reg req'd) is doesn't delve into these externalities. Also, the stats don't seem that impressive. Or maybe malpractice is complicated and multidimensional.

A more appropriate medical application of the Blink thesis is made regarding Goldman's Chest Pain algorithm (free registration necessary). Gladwell cites Goldman's conclusion -- that the likelihood of chest pain being due to heart attack is largely dependent on just three pieces of data -- as an example of "less is more" -- it turns out that Goldman's criteria can ferret out MIs better than docs with tons of data on patient age, heart rate, appearance, etc.

I'm all for data mining to determine which test results are truly diagnostic. And I'm always impressed with seeming mind-readers (Blink goes into detail about what should be called face-reading) who can divine useful information from flickers of facial muscle contractions. As far as medicolegal studies are concerned, I don't think we're at the point where expressions and intonations can coexist alongside assays and echocardiograms as actionable data, subject to statistical interpretation. That's just my snap judgment.