Wednesday, July 11, 2007

Hi-Tech fMRI lie detection - another scam?

Margaret Talbot writes an interesting article in The New Yorker (PDF here) about the current hype over lie detection using brain scanning techniques, and the sprouting of numerous companies wanting to sell their services to private investigators, police departments, U.S. and foreign government agencies, etc. The article is a broad discussion of lie detection also by other physiological indicators such as skin conductance and blood pressure. Most interesting to me were the critical points raised. Here are a few clips from the article:
Paul Bloom, a cognitive psychologist at Yale, believes that brain imaging has a beguiling appeal beyond its actual power to explain mental and emotional states. "Psychologists can be heard grousing that the only way to publish in Science or Nature is with pretty color pictures of the brain," he wrote in an essay for the magazine Seed. "Critical funding decisions, precious column inches, tenure posts, science credibility, and the popular imagination have all been influenced by fMRI's seductive but deceptive grasp on our attentions." Indeed, in the past decade, Nature alone has published nearly a hundred articles involving fMRI scans. The technology is a remarkable tool for exploring the brain, and may one day help scientists understand much more about cognition and emotion. But enthusiasm for brain scans leads people to overestimate the accuracy with which they can pinpoint the sources of complex things like love or altruism, let alone explain them.

Brain scans enthrall us, in part, because they seem more like "real" science than those elaborate deductive experiments that so many psychologists perform. In the same way that an X-ray confirms a bone fissure, a brain scan seems to offer an objective measure of mental activity. And, as Bloom writes, fMRI research "has all the trappings of work with great lab-cred: big, expensive, and potentially dangerous machines, hospitals and medical centers, and a lot of people in white coats."

Deena Skolnick Weisberg, a graduate student at Yale, has conducted a clever study, to be published in the Journal of Cognitive Neuroscience, which points to the outsized glamour of brain-scan research. She and her colleagues provided three groups--neuroscientists, neuroscience students, and ordinary adults--with explanations for common psychological phenomena (such as the tendency to assume that other people know the same things we do). Some of these explanations were crafted to be bad. Weisberg found that all three groups were adept at identifying the bad explanations, except when she inserted the words "Brain scans indicate." Then the students and the regular adults became notably less discerning. Weisberg and her colleagues conclude, "People seem all too ready to accept explanations that allude to neuroscience."

Nancy Kanwisher, a cognitive scientist at M.I.T., relies a great deal on MRI technology. In 1997, she identified an area near the bottom of the brain that is specifically involved in perceiving faces. She has become a pointed critic of the rush to commercialize brain imaging for lie detection, and believes that it's an exaggeration even to say that research into the subject is "preliminary." The tests that have been done, she argues, don't really look at lying. "Making a false response when instructed to do so is not a lie," she says. The ninety-per-cent "accuracy" ascribed to fMRI lie detection refers to a scenario so artificial that it is nearly meaningless. To know whether the technology works, she believes, "you'd have to test it on people whose guilt or innocence hasn't yet been determined, who believe the scan will reveal their guilt or innocence, and whose guilt or innocence can be established by other means afterward." In other words, you'd have to run a legal version of a clinical trial, using real suspects instead of volunteers.

She points out that the various brain regions that appear to be significantly active during lying are "famous for being activated in a wide range of different conditions--for almost any cognitive task that is more difficult than an easier task." She therefore believes that fMRI lie detection would be vulnerable to countermeasures--performing arithmetic in your head, reciting poetry--that involve concerted cognitive effort. Moreover, the regions that allegedly make up the brain's "lying module" aren't that small...Saying 'You have activation in the anterior cingulate' is like saying 'You have activation in Massachusetts.' "

Kanwisher's complaint suggests that fMRI technology, when used cavalierly, harks back to two pseudosciences of the eighteenth and nineteenth centuries: physiognomy and phrenology. Physiognomy held that a person's character was manifest in his facial features; phrenology held that truth lay in the bumps on one's skull. In 1807, Hegel published a critique of physiognomy and phrenology in "The Phenomenology of Spirit." In that work, as the philosopher Alasdair MacIntyre writes, Hegel observes that "the rules that we use in everyday life in interpreting facial expression are highly fallible." (A friend who frowns throughout your piano recital might explain that he was actually fuming over an argument with his wife.) Much of what Hegel had to say about physiognomy applies to modern attempts at mind reading.

Elizabeth Phelps, a prominent cognitive neuroscientist at N.Y.U., who studies emotion and the brain, questions another basic assumption behind all lie-detection schemes--that telling a falsehood creates conflict within the liar. With the polygraph, the assumption is that the conflict is emotional: the liar feels guilty or anxious, and these feelings produce a measurable physiological response. With brain imaging, the assumption is that the conflict is cognitive:the liar has to work a little harder to make up a story, or even to stop himself from telling the truth. Neither is necessarily right. "Sociopaths don't feel the same conflict when they lie," Phelps says. "The regions of the brain that might be involved if you have to inhibit a response may not be the same when you're a sociopath, or autistic, or maybe just strange. Whether it's an emotional or a cognitive conflict you're supposed to be exhibiting, there's no reason to assume that your response wouldn't vary depending on what your personal tendencies are--on who you are."

No comments:

Post a Comment