Saturday, February 4, 2012

Lies Scientific American Told Me: A Response to Michael Shermer's "Lies We Tell Ourselves"

First, let me say that I generally really enjoy Scientific American articles. They introduce me to new topics, new perspectives, and new writers. They make me think (and I suppose this article I'm about to rip apart also made me think).

I was really disappointed, however, with a recent article by Michael Shermer about Robert Trivers's book, The Folly of Fools, which outlines why human beings evolved the ability to not only be deceptive but to detect deception in others. This theory draws upon evolutionary theory as well as game theory - a theory about how people make strategic decisions when making exchanges with others (and how we use information on others' strategies to influence our own strategy).

The problems with Shermer's article have been well-outlined by the commenters. I urge you to read them, because they are generally articulate and thoughtful - I didn't come across any that appeared to be people "trolling" (see previous blog post for more on Internet trolls). Instead of repeating what these commenters already said so eloquently, I want to focus instead on this issue of lie detection.

Trivers argues that when people are being deceptive, they have three big "tells": nervousness, control (to hide their feelings of nervousness), and cognitive load (which results from having to construct a whole new reality, and has several noticeable effects on behavior, such as fewer hand gestures and speaking in higher-pitches).

People have believed in these tells for a long time (and the media continues to propagate these and similar beliefs). This is why people administering the polygraph and other "lie-detecting instruments" include control questions - questions that people generally answer truthfully - to provide a baseline to which we can compare responses to the questions of interest. If, while lying, they do things like pause longer or speak in higher tones, then voice stress analysis units should be effective at detecting lies, right? And yet, in research, VSA is not found to be effective at differentiating liars from those telling the truth (read a brief summary here), prompting one judge to refer to a particular VSA device as "little better than a sewing machine".

Also doubles as a paper-weight
So perhaps this means we shouldn't depend on computers and other electronic devices. After all, according to Trivers, human have evolved many of these capabilities, and can examine really complex things, like context, that even the smartest computer can't handle. We should instead use people to detect lies. And yet, people are generally very poor at detecting lies on their own (an unpublished meta-analysis mentioned in this article found overall detection deception of 53%, little better than chance). Whenever I teach Psychology & Law, I also do a deception detection activity, where students receive a card telling them to lie or tell the truth to a partner who asks them "What did you do yesterday?" Except for one class that was very good at detecting lies (something like 70%), every class scores right around 50% - chance.

So, as the article linked just above discusses, people have responded by introducing training, especially for people who detect deception as part of their job: police officers, airport security, and so on. That should help, right? A study by Kassin and Fong (1999 - full article here) found that training actually reduced accuracy in detecting deception, but increased confidence. Those who received training were more likely to be wrong than those who did not receive any training, but were more confident that they were right.

This cavalier belief that we can detect deception, and can only improve with greater training, has led to many programs and initiatives meant to bring better deception detectors to places like the airport. And that feeling of confidence in response to training only reinforces the belief that training is a good idea and should be carried out for more personnel. In response to the terrorist attacks of 9/11, the United States has implemented different programs that provide training to airport security agents to detect when people are hiding something, whether that be drugs, money, or explosives.

One program utilizes micro-expressions. These flickers (lasting about 1/25th of a second) of emotion appear on our faces before we have conscious control over our outward expressions of emotion. These micro-expressions have been well-studied by Dr. Paul Ekman, an expert in human emotion. Some of Dr. Ekman's first research on the issue established the 6 universal human emotions - happiness, sadness, anger, fear, surprise, and disgust - that are present across cultures. Based on his research on human emotion generally and micro-expressions specifically, he has developed short training (delivered in as little as 30 minutes).

In as little as 30 minutes, we've convinced TSA agents that, while carrying out various other parts of their jobs, they can catch terrorists.

Dr. Ekman is a brilliant man who has been studying human emotion since the 1950s. This man can probably read any emotional expression, no matter how short, because he knows so much about it. Expecting TSA employees receiving brief training to be able to use these micro-expressions to detect deception at even half the level of someone like Paul Ekman is, I think, expecting a lot.

Do I think these micro-expressions exist and provide cues about lying? Absolutely. But do I think that one can go through brief training, where the micro-expressions are either pre-recorded or, worse yet, acted out, and then walk into the real world, without control or scripts (or feedback on most of the decisions they make about passengers' deception) and use this tactic effectively? Absolutely not.

Remember, in general (unless the airport is performing extra searches on people completely at random) the only people who are stopped and inspected are the ones the TSA agent suspects of being deceptive. This leaves a bunch of people not believed to be deceptive walking right through security, and no feedback on whether they actually were hiding something.

I might be more comfortable with a program that simply clones multiple copies of Paul Ekman to run around our major airports catching liars. Of course, then they'd all want their own TV show.

Thoughtfully yours,