The Liar's Tell: Is Paul Ekman Stretching the Truth?

"Ekman now admits it was a mistake to define micro-expressions by duration, and describes them now as any facial expression that is very difficult to detect in real time."

I can see why he regrets having made it--it's the only concrete, measurable claim he made (that's mentioned in the article).

I'ts been said before, but there's a major issue with any program that depends on TSA officers' perception of people as liars or truthful: true positives are going to be swamped by officer bias. In lower-stakes situations study after study shows police officers presuming non-white people to be more likely to be criminals without regard for local statistics. A terrorist group probably has a higher chance of slipping a member by who is or can pass for non-Arab under this setup. I doubt most domestic terrorists (almost universally white) are going to get the same scrutiny no matter how nervous as a frustrated Arab-American woman from Des Moines will. At best they might get checked in case they were smuggling.

The model here relies on teaching officers to select people based on whether they think they're lying, it's a perfect situation for prejudice to distract from actual threats. Bruce Schneier and others have pointed how this wastes resources and increases the attack surface for terrorists, but the TSA has been highly resistant to feedback & criticism. It's the equivalent of having the physical security guys for a building do sweeps at night to look for post-its with passwords. That's not the computer security threat and it's not the breaking & entering one, so why do you have them wandering around in the dark?

Anxiety occurs for multiple reasons, so false positives would always be a confounding factor, even if micro expressions occured 100% of the time.

There is quite a bit of evidence that even with physiological measures linked to stress and lie detector machines, there is little reason to believe lie detectors are especially effective. Given that non invasive means of detecting anxiety are limited, and anxiety is itself limited, such screening is going to not be helpful.

That said, there are some great studies on using recognition and EEGs or fmri machines to detect lies. Your brain processes item's you recognize in different ways, so if someone maintains they never were in an apartment where a murder occured nor had seen it ever, you could display a series of apartments, including the murder scene, and be able to tell when they recognized the apartment. It's known in the field as the Guilty Knowledge Test, or GKT. Check out the work by Farwell and others if you are interested.

I don't know if Ekman is real or not (probably not), but any serious study of predictions should report both precision (are your accused liars all guilty?) and recall (did you catch all the liars?), not make statements like "70-percent accuracy." Especially if we are using these techniques for security screening, these numbers are essential. We'll also need to educate security personnel about the unintuitive nature of false positives, where even if a test achieves a low false positive rate, a given flagged person still has a low chance of being a liar (see http://en.wikipedia.org/wiki/False_positive_paradox).

Ekman is no less of of voodoo priest than polygraphers - an entire so-call profession of witch doctors. He has cashed-in magnificently on a government and population willing to buy empty high-tech looking boxes from a dowsing practitioner to detect bombs.

It would be interesting to read an objective analysis of Israel's airport security screening, to compare with what Ekman and the TSA have concocted. There are plenty of glowing articles [1] about the Israeli "gold standard", but I'm sure there are also plenty of glowing articles about Ekman.

It's certainly a field with many shades of gray. Not sure if most on HN can relate to any of this, engineers are mostly about facts, about black and white. Emotions don't have a lot to do with coding or debugging (except for frustration, no shortage of that!).

[1] http://www.huffingtonpost.com/daniel-wagner/what-israeli-air...