This post will, in all likelihood, be very brief and to the point. It isn't so much that I've grown weary of my often overly-verbose descriptions of the simplest phenomena. Rather, this topic is so self-explanatory and the conclusions so self-evident that (I would hope) no further discussion will be warranted.
As you may have easily noticed, there is a rather cliche' image above of an individual hooked up to a polygraph machine. I will spend the next x-amount of sentences decrying the use of these contraptions. Let's get this show on the road:
1) Allow me to first direct your attention toward one of the components seen in the picture with which I am quite familiar: the "Galvanometers." These are electrodes situated on the undersides of the fingers, and measure Galvanic Skin Response (GSR), otherwise known as Skin Conductance Response (SCR). I'll use the latter term. Here's the thing about SCR: if you tell a blatant lie, you're aware that you're lying, and realize that the truth could get you in trouble, you will probably show an observable response. Know what else could cause such a response? Being on trial for murder and knowing that your answer to each question and the readings accompanying them may determine whether you're killed or acquitted. The truth of the matter is that there are many, many things that can produce what appears to be a skin conductance response, including but not limited to deep breaths and thinking about the person in front of you naked. These are difficult enough to tease out when in response to a discrete event in a tightly controlled experiment.
2) Next, I'll simply lump blood pressure, respiration (pneumography), and heart-rate into the same category since my criticisms can by and large be levied in conjunction. Quite like SCR, these measures are, without proper context, rather nonspecific. These, however, suffer even more from such ambiguity. There's always a difficulty in making meaningful comparisons within this sort of transient activity without establishing a proper baseline, which I would be willing to bet many polygraph operators don't bother to do. What you're left with is an array of autonomic responses that, again, are probably highly influenced by the fact that these people have been accused of something like murder, and may potentially be put on death row. To ask them to stay calm for good readings would be like asking a hungry polar bear not to eat you so you can see who wins the World Cup. When faced with a moment or time frame that may determine life or death, you effectively have no baseline, and your readings are inherently biased. Particularly because the responses in question (including SCR) are tightly coupled with nonspecific stress. An additional concern in using this method to investigate homicides is that many sociopaths have been shown to have weaker or relatively nonexistent autonomic responses. What does this mean? Well, in a lot of cases, it means that you may only be capable of catching murder suspects lying who didn't actually commit the murder.
3) Here's the main event. Ready for it? Here it is. Ever heard of the "black box" problem? If so, you've probably already got the idea. If not, allow me to briefly explain. Most polygraph tests simultaneously record three or four different measures within the same physical object. Not three or four distinct pieces of hardware with three or four distinct methods of amplifying and cleaning up the signal: the same piece of hardware. These three or four signals go into one machine that assigns parameters to each measurement, and spits them out. You may ask yourself why this is a problem if each parameter is assigned a distinct waveform area on the paper/computer/etc...well, here's the problem. The "black box" concept refers to the idea that data are being thrown into a program or machine that manipulates the data in set ways and spits it back out in a way that is defined as more easy to interpret or compare. However, without adequate knowledge of what's going on inside the program or machine (the "black box"), you're really not aware of where the numbers or waveforms are coming from. That is, if you don't know how the input becomes the output, how can you know that the transformation was accurate? The machine could easily malfunction, and since all of the readings may be affected, you would probably never know. This problem is only worsened when all of the readings are combined into a single waveform, which is a shit idea for the aforementioned reason, and furthermore because something like SCR and blood pressure are only loosely related and totally incompatable within the same frame of measurement. This may be somewhat less problematic if the operators/analysts of polygraph equpment had an extensive theoretical and technical knowledge of the measurements and algorithms involved, but I would be willing to bet that this very often isn't the case.
Thanks for reading.
No comments:
Post a Comment