The New Scientist asks "Why waste money on technology when humans can still do better?" and then offers an answer. Here's part of its response.
IF YOU want to spot a liar, don't bother with a polygraph. They are notoriously unreliable. In a competition to find the world's most inappropriately named technology, the lie detector would be hard to beat.
Yet lie detectors are still relied upon to a frightening extent. Law enforcement agencies in the US, Canada, Israel and elsewhere use them routinely in investigations, even though evidence based on lie detectors is usually inadmissible in court.
Dissatisfaction with traditional polygraphs has prompted the search for more sophisticated techniques to spot the supposed telltale signs of dishonesty - increased heart rate, sweating, nervous tics, averted gaze and so on. The US Department of Homeland Security, for example, is spending millions on asystem to detect "malintent" in airline passengers to weed out potential terrorists.
And yet the assumption that anxiety causes liars to give themselves away is dubious. A 2003 review of the evidence by the US National Academy of Sciences concluded that while lie detectors are significantly better than a toss of a coin, they fall far short of the accuracy required. Little has changed.
Why do we keep using lie detectors? I think part of the reason is that we want that finality--we want certainty to the question--is someone telling the truth? Remember that old game show, To Tell the Truth? We want to know the answer. Most of us don't like ambiguity or uncertainty. We seek out solutions, for the same reasons that we seek out certainty in other areas of our lives. Many of us seek the comfort of religion. We look for patterns in clouds. Many of us carry good luck charms such as rabbits' feet and visit fortune tellers to find out our futures. We cling to the inserts in Chinese fortune cookies and eye the newspaper horoscopes. Most of us wouldn't like a mystery story that ended without a solution. One episode of the tv series M.A.S.H., "The Light That Failed," centered on the characters' frantic search for the solution to a mystery story once they discovered that the last page is missing. In desperation, B.J. finally calls the author to find out "Whodunnit."
When faced with the responsibility of serving on a jury panel, we'd like the same kind of certainty, coupled with scientific credibility, and a magic machine that one could hook up to an individual to detect truth or falsity might seem to give us that. Whether it's a polygraph (the results of which are usually not admissible) or some other magical seeming device (whose accuracy is still not proven) it's probably got wires and a gazinta--it gazinta the wall. It will come with someone in a metaphorical lab coat to explain its meaning. What more could one ask? The law and neuroscientists seem to trying to move in that direction. Well, one could ask, nay demand, for some more actual science to go along with the wires and the gazinta. If someone's life depends on the certainty, the science had better be certain. In spite of the wires and the gazinta.
Any law student who has taken Evidence has read about, or better experienced, an experiment in which a man bursts into a crowded classroom, runs through shouting and then leaves. When questioned directly after the event there is strong disagreement among the witnesses as to what the man was saying, what he was wearing and whether or not he had a gun. Based on the work of psychologist Elizabeth Loftus, now on the faculty of the University of California at Irvine Law School, this experience, more than any dry article about cognitive science, demonstrates the inherent unreliability of human memory and the conviction of eye-witnesses about what they have seen. Lawyers involved in the Innocence Project which is seeking to challenge wrongful convictions based on eye-witness testimony by examining conflicting DNA evidence have further brought these findings to public attention. As they explain, "Research shows that the human mind is not like a tape recorder; we neither record events exactly as we see them, nor recall them like a tape that has been rewound." Yet despite what has become common knowledge about the malleability of human memory, the idea that it’s possible to access the brain directly to find out whether a witness is telling the truth is being put forward by companies which seek to profit from research that suggests that new imaging technology can detect when a human is telling a lie. These companies are advertising this technology as a tool for law enforcement and promoting its use in U.S. trials as a way of helping juries to assess the credibility of witnesses.
This article explores these claims that neuroimaging scans can be used to detect lies, which far exceed those made by responsible scientists, and also puts them in the context of a series of U.S. Supreme Court cases which have dramatically changed how scientific (forensic) evidence can be presented to the jury in criminal trials. See Daubert v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579, 113 S.Ct. 2786, 125 L. Ed. 2d 469 (1993)(establishing new criteria for admission of scientific evidence); Crawford v. Washington, 541 U.S. 36 (2006) (requiring that defendants directly face accusors). It also addresses the significant criticisms being brought against what has often been incautious adoption of unreliable techniques. 'Strengthening Forensic Science in the United States: A Path Forward' (National Research Council 2009).
In this article I argue that promises of lie detection are not only based on false premises, but they are harmful to the integrity of the legal system because they seek to substitute a technology, which is not just undeveloped and inadequately tested but inherently flawed, for the judgment of the fact-finder, judge or jury, in a criminal trial. I conclude that even if there was neuroimaging technology which could provide direct access to human thought, the result would share the inaccuracies and subjectivity that we already know is an inherent feature of human memory. Moreover, because this technology promises to do something that jurors know they cannot - determine when a person is lying - there is a substantial risk that it will prejudice defendants because jurors will substitute the results of the technology for their own collective judgment.
Despite media hype and at least two companies marketing themselves as offering scientific expert testimony admissible in US courts on whether a witness is telling the truth, recent decisions by a trial court in New York and a Federal Magistrate Judge in Tennessee indicate that replacing the jury deliberation room with a magnetic resonance imaging machine is not in the foreseeable future.
Functional magnetic resonance imaging (fMRI), an emerging field of neuroscience that seeks to correlate brain activity to behavior, has been touted by some to be the new lie detector. The MRI machine is able to “see” the increase in oxygenated blood as it fuels various regions of the brain during actions and activities. The hypothesis underlying fMRI as a lie detector is that telling the truth is the natural or normal response of the brain and one would not expect to see increased activity over and above the normal background level of brain activity. But when the subject begins to prevaricate, more brain activity is needed and more oxygenated blood is directed to those regions of the brain processing the “lie.” This blood oxygenation level-dependent (BOLD) differential is measurable by the fMRI.
Comparing the BOLD differential between subjects known to be telling the truth and then deliberately lying allows researchers to hypothesize that increased BOLD above the base line in certain regions of the brain when the subject is answering questions is an indication of deception.
It might be in the lab, but the rules of evidence as applied in US courts considers the citizens of the community selected to sit on the jury more reliable “lie detectors” than the new technology represented by fMRI. Two recent decisions, one in a Federal court applying the Federal Daubert standard and one is New York state court applying the Frye general acceptance test, rejected expert testimony based on the results of fMRI scanning.
In finding fMRI failed to satisfy the exacting standard of Federal Rule of Evidence 702, a Federal magistrate judge in United States v. Semrau measured the new technology against the Daubert reliability factors. fMRI passed the first two factors, testing and peer review. The court noted that the technology was certainly capable of and was being tested and that numerous articles concerning fMRI had appeared in the peer reviewed scientific literature.
The next two Daubert factors, the known or potential error rate and the existence and maintenance of standards, however posed significant problems. While there may be some identifiable error rates associated with fMRI in the controlled lab setting, it was undisputed that error rates in the “real world” application are completely unknown. Similarly, protocols exist for lab studies of fMRI but not such standards exist for real-life applications. In fact, Dr. Steven Larkin, the expert witness for plaintiff in Semrau, and one of the leading proponents of fMRI as a lie detector admitted to deviating from his own protocol in administering the test on Dr. Semrau.
Finally, the court found that fMRI as a reliable lie detector had not achieved general acceptance by the scientific community, the fifth Daubert factor. By failing to meet three of the five Daubert factors, the magistrate judge concluded that fMRI was not sufficiently reliable to admit under FRE 702.
Download the paper from SSRN at the link.
The magistrate judge also found the use of fMRI to bolster Dr. Semrau’s credibility more prejudicial than probative. Relying primarily on cases excluding polygraph examinations taken without the knowledge or participation of the opposing party, the magistrate judge concluded that the unilateral fMRI test here lacked probative value because Dr. Semrau faced no negative consequences by undergoing the test. Accordingly, the evidence was excluded under FRE 403.
The New York case, Wilson v. Corestaff Services L.P., applied the Frye general acceptance standard to fMRI based expert testimony to bolster the plaintiff’s witness in an employment discrimination case. The court was troubled be the notion that fMRI based expert testimony would usurp the jury’s responsibility to determine the credibility of the witnesses, a dramatic departure from the traditions of a common law trial. Additionally, the court found the plaintiff was unable to show the use of fMRI to determine truthfulness had been generally accepted by the relevant scientific community, the standard for admissibility under Frye.
The analysis in Semrau and Wilson combined imposes significant hurdles to the admissibility of fMRI on the issue of witness credibility. The procedure failed to meet the reliability standards of FRE 702 and the general acceptance requirement of Frye. Furthermore, concerns over jury confusion, usurpation of the traditional function of the jury in judging credibility, and whether such testimony is even an appropriate matter for expert opinion raise additional hurdles. For proponents of fMRI expert testimony on witness credibility the obstacles to admissibility are significant under current law.