Thermal Imaging "Lie Detector"

[George W. Maschke]

The Mayo Clinic of Rochester, Minnesota has posted a news release titled New Technology Detects Lying, Paves Way for Increased Security on attempts to detect lying through the use of thermal imaging technology. This research was led by Dr. James Levine of the Mayo Clinic with the participation of Norman L. Eberhardt, also of the Mayo Clinic, and Ioannis Pavlidis of Honeywell Laboratories.

The Mayo Clinic website also includes a link to a second page with video and audio files about Dr. Levine's research.

A "brief communication" about this research was published in the 3 January 2002 issue of Nature under the title, "Human behaviour: Seeing through the face of deception."

It seems to me that Dr. Levine has drawn conclusions that go far beyond the evidence of his research. For example, his thermal imaging camera did not detect "lying," as the Mayo Clinic news release claims. Rather, lying was inferred based on increased temperature of the eye sockets. This technique seems likely to be beset with the same problems as polygraphy.

[beech trees]

But, George. . .

We know this new technology is accurate because they verified their results with polygraphs!  

[G Scalabr]

Yes, this device does indeed seem to be plagued with the same problems as polygraphy. Most importantly, the "base rate" problem would make it completely useless in an airport setting.

I'm going to be generous and estimate that the base rate of terrorists to legitimate passengers is 1 per 1,000,000. The UPI article reports that the device is 80 percent accurate. This means that the "test" is 80/20 (percentages) with liars and 80/20 with truthful people. When that one terrorist comes through, there will be an 80% chance that he will set it off. On the other hand, 200,000 truthful people will be falsely accused. So for every terrorist that comes through, we still have to pick him out of the 200,000 innocent people who have also set the device off. Thus, the chance of catching a terrorist with this test (its predictive validity) would be .0005 (or 5/1000 of a percent). On the other hand, I could simply declare everyone truthful and my prediction would be 99.9995 accurate.

As Fred F said in another thread, all this device would do in airports is create chaos.

[George W. Maschke]

Bob Park of the American Physical Society writes the following in the 4 January 2002 installment of his What's New column:

3. AIRPORT LIE DETECTOR: AT FIRST BLUSH, IT'S A DUMB IDEA.
The system is supposed to scan the faces of passengers at the check- in counter with a high-definition thermal imaging camera while they answer questions. The claim is that blood rushes to the eye area when people lie. That may be, but your face will also flush when you run three miles from Concourse A to Concourse F, only to find the gate has been changed. On the other hand, you may turn pale if you see fuses dangling from another passenger's shoes. In short, it will work as well as a polygraph, which is not at all.