Truth Machine

I find polygraphs fascinating. The idea of using a machine to exploit bugs in human behavior to discern objective truth from falsehood is just an irresistible notion to a story-minded person like me. To have a machine that can cut through the illusions and deceptions of human stories is just so metaphorically resonant. Of course, I know that polygraphs aren’t really lie detectors, not in the way they’re imagined. At best they monitor a person for signs of physiological stress as a reaction to making up lies on the spot. This is easily lost in background noise, and easily sidestepped by rehearsing a convincing lie ahead of time. 

A large part of the machine’s job is to make a subject afraid to lie in the first place, which makes lies easier to spot. It doesn’t work if the subject believes the lie, or doesn’t experience stress while telling it, nor is it effective on people who fall outside of some basic stereotypes about liars. Eye surgery, heart arrhythmia, brain damage, and ambidextrousness can all throw a polygraph to the point of uselessness. At worst, polygraphs provide a prop for interrogators to confirm their own biases and coerce a subject into believing they’re trapped, whether or not they’re actually guilty, or else to convince jurors of an unproven circumstantial case. 

Still, they’re fascinating. The kabuki theater act that interrogators put on to try and maneuver the subject into the correct state of mind to find a chink in the psychological armor, the different tactics, the mix of science and showmanship is exciting to explore. I enjoy reading through things like polygraph manuals, and the list of questions used in interviews of federal employees for security clearance. 

What’s interesting is that most of the questions are just bad. Questions like “Prior to [date], did you ever do anything dishonest?” are just bad questions. After all, who decides dishonesty? Is a dishonest act only an action committed in service of a direct, intentional lie, or is it broader? Does omission count as an act in this context? Is dishonesty assessed at the time of the act, or in retrospect? Would a knowing deception made in the interest of a unambiguously moral end (for example, misdirecting a friend about a Christmas present) constitute a dishonest act? 

These questions are listed in the manual as “No-answer Comparison Questions”, which if I understand the protocol correctly, are supposed to be set up such that a subject will always answer “No”, and most of the time, will be lying. The idea here is to establish a baseline, to get an idea of what the subject looks like when lying. The manual suggests that these questions will always be answered with “no” because, earlier in the interrogation, the interrogator will have made clear that it is crucial for subjects to provide an impression of being truthful people. The government, the interrogator is instructed to say, doesn’t want to work with people who lie or cheat, and so it is very important that people going through this process appear honest and straight laced. 

Of course, this is hogwash. The government does want people who lie, and it wants people who are talented at it. A general needs to be talented at deception. An intelligence operative needs to keep secrets. Any public figure dealing with sensitive information needs to be able to spin and bend the truth when national security demands it. Even the most morally absolutist, pro-transparency fiend understands that certain government functions require discretion with the truth, and these are exactly the kind of jobs that would involve polygraph tests beforehand. 

The government’s polygraph interrogation protocols rely on subjects swallowing this lie, that they need to keep a consistent and presentable story at the expense of telling the truth. They also rely on the subject recognizing that they are lying and having a reaction, since a polygraph cannot in itself divine material truths, but work only by studying reactions. For it to really work, the subject must also be nervous about lying. This too is set up ahead of time; interrogators are instructed to explain that lying is a conscious and deliberate act, which inspires involuntary physiological fear in the subject. This is arguably half true, but mostly it sets up a self-fulfilling prophecy in the mind of the subject. 

It’s pretty clear that the modern polygraph is not a lie detector. But then again, how could it be? Humans can barely even agree on a consistent definition of a lie within the same language and culture. Most often we tie in our definition of lying with our notions of morality. If you used deception and misrepresentation to do a bad thing, then you lied. If you said something that wasn’t true, but meant nothing by it, and nothing bad came out of it, well then you were probably just mistaken. I don’t want to make this post political, but this trend is obvious if you look at politics: The other side lies, because their ranks are filled with lying liars. By contrast, our side occasionally misspeaks, or is misinterpreted.

This isn’t to say that there’s no such thing as truth or lies, just that we can’t seem to pin down a categorical definition, which you do need if you’re going to program a machine to identify them. We could look for physiological reactions involved in what we collectively call lying, which is what polygraphs purport to do, but this just kicks the problem back a step. After all, what if I genuinely and wholeheartedly don’t consider my tactful omission about “clandestine, secret, unauthorized contact with a non-U.S. citizen or someone (U.S. citizen or non-U.S. citizen) who represents a foreign government, power, group or organization, which could result in a potential or real adverse impact on U.S. national security, or else could result in the unauthorized aid to a foreign government, power, group or organization” to be a lie? If the machine is testing my reactions, it would find nothing, provided I didn’t believe I had anything to lie about. 

This is where competent question design and interrogation technique is supposed to obviate this issue. So, a competent interrogator would be sure to explain the definition of contact, and foreign power, and so on, in such a way that would cause me to doubt any misconceptions, and hopefully if I’m lying, trigger a stress reaction. The interrogator might insinuate that I’m withholding information in order to get me to open up, or try and frame the discussion in such a way that I would think opening up was my only option. But at that point, we’re not really talking about a lie detecting machine, so much as a machine that gives an interrogator data to know when to press psychological attacks. The main function of the machine is to give the interrogator certainty and undermine my own confidence, so that the interrogator can pull off bluffing me into cracking. 

So are polygraphs useful? Obviously, as a psychological tool in an inquisitional interrogation, they provide a powerful weapon. But are they still more useful than, say, a metal box with a colander attached? Probably, under some circumstances, in the hands of someone familiar with the underlying principles and moving parts of both psychology, physiology, and the machine itself. After all, I don’t think there would be such a market if they were complete bunk. But then again, do I trust that they’re currently being used that way by the groups that employ them? Probably not.

Works Consulted

Burney, Nathan. “Convict Yourself.” The Illustrated Guide to Law, lawcomic.net/guide/?p=2494.

United States, Department of Defense, Polygraph Institute. “Law Enforcement Pre-Employment Test.” Law Enforcement Pre-Employment Test. antipolygraph.org/documents/dodpi-lepet.pdf.