July 16, 2010

To catch a liar: Don't watch Fox-TV

When the TV show Lie to Me jumped off in 2009, the hype presented it as grounded in true-to-life science of lie detection. Social scientist Paul Ekman of the University of California at San Francisco, upon whom the show is based, even critiques the scientific accuracy of each episode on his Fox-TV blog.

But watching the show actually makes people WORSE at detecting deception, while at the same time increasing their overall suspiciousness and cynicism about others' honesty, according to a carefully designed study just published in the journal Communication Research.

"Lie to Me appears to increase skepticism at the cost of accuracy,” reports the research team led by Timothy Levine of Michigan State University.

As reported by Tom Jacobs at Miller-McCune, the findings have real-world implications:
Levine and his colleagues argue that … most recent research casts doubt on the accuracy and effectiveness of lie-detection methods presented on the series as unfailingly successful…. So once again, "fictional media portrayal of social science theory leads to confusion between fiction and fact," the researchers write. "Viewers (of the show) may come away with the false sense they can better detect lies. Viewers may also acquire a false sense that law enforcement officers are being effectively trained to detect deception and, therefore, may be less critical as jurors or witnesses." So the next time you turn on a television show, keep in mind that the creators just may be lying to you.
Jacob’s full report on the study is HERE.

Related blog posts:


DM said...

Without having seen the actual statistics, I'm wondering whether the difference in detection accuracy between the Lie to Me sample and the control sample was greater than a standard deviation. The figures reported by Jacobs for the Miller-McCune are awfully close. Furthermore, watching Lie to Me didn't lower people's lie-detecting abilities below the threshold of chance; they could still detect whether a person was telling the truth or fibbing more than 50% of the time.

I dunno. I watch Lie to Me, and I like that there are tie-ins with Ekman's research, but I always take the assertions they make with a grain of salt and engage a little suspension of disbelief. I don't think this research is going to keep me from watching.

dredslack said...

Let's differentiate between fiction and reality: Lie to Me is a TV crime drama not the reenactment of the proven techniques of a truth divining virtuoso. Interesting story lines, characters and settings are there to entertain and sell commercial time not increasing the audiences lie-detecting abilities.
I really enjoy the show but, c'mon, scientific reality? Hardly. Roth deftly plays his nearly unhinged, irreverent character for all he's worth (and looks to be having a damn fine time as he's at it). It's brilliant theatre with just enough science sketched in to keep our minds engaged, but it's not serious science. No one could develop a baseline that quickly and account for variables such as culture, experience, level of training, intelligence etc., etc. Furthermore, humans are not robots and will react to stress differently ... so the 'tells' Dr. Lightman uses are going to change a lot between subjects, where on the show they all mean pretty much the same.
It would be easy to go on and pick the show apart for it's scientific inaccuracies but all that really tells us what we already know, it's a TV show!
A much more useful purpose for all this brain power would be to get a ways and means going to teach the general populace a few things about the techniques that are used to lie to them each day. Misassociation, disproportioning, burlesquing, baiting and misdirection are just a few of the simple techniques that are used consistently to lie to and manipulate people. Sadly, many are completely unaware.
Perhaps a desire to learn how to not be deceived is part of the shows popularity. Certainly that information is available but it won't be found on Lie to Me.
I'll still be watching

Matthew said...

So it's about as accurate as a lie-detector machine. You're still leaving your life in the hands of someone who may, or may not, have the ability to decipher the truth from lies. Atleast the Lie To Me techniques aren't used in the real world in order to convict someone of a crime they may have done.

You should also look a bit deeper into Lie To Me. The show doesn't try to use the techniques to convict someone solely on saying their lieing. They do it to get a confession. So weither it's real or not, it would get the job done better than a lie detector in real life.

Anonymous said...

Table 1. Means (and Standard Deviations) for Truth Bias and Accuracy by Experimental Condition

Dep Variable Lie to Me Numb3rs Control

Judged True 50.7% (13.4) 57.9%(18.1) 59.0% ( 15.8)
Tot accuracy 59.5% (12.1) 61.7% (13.8) 65.2%(11.7)
Truth accuracy 60.1% (19.1) 69.6% (21.6) 74.3% (16.3)
Lie accuracy 58.8% (17.0) 53.8% (24.0) 56.2% (22.5)

Anonymous said...

Examining differences among conditions, the omnibus ANOVA for truth bias approached statistical significance, F(2, 105) = 2.71, p = .071. A planned contrast indicated that the combined means of the control and Numb3rs groups (M = 59.0% and 57.73%, respectively) were significantly greater than the mean in the Lie to Me group, M = 50.7%, t(105) = 2.32, p = .023. Further, a Dunnett test shows that truth bias in the Lie to Me group was significantly lower than the control, p = .032.
Neither the omnibus ANOVA for overall accuracy, F(2, 105) = 1.79, p = .17, η2 = .03, nor for lie accuracy were statistically significant, F(2, 105) = 0.49, p = .61, η2 = .00. There was, however, a significant difference for truth accuracy, F(2, 105) = 4.76, p = .01, η2 = .08, r = .29. Tukey B and Dunnett’s post hoc tests showed lower truth accuracy in the Lie to Me group than in the no show control.