Post No.: 0521
Furrywisepuppy says:
The accuracy and validity of polygraph tests are dubious, yet some public and private sector organisations still rely on them as interrogation tools, such as the probation service or as part of workplace assessments. Polygraph, or ‘lie detector’, machines might measure a subject’s galvanic skin response (sweating), blood pressure, heart rate, respiration rate and/or movements (microexpressions or amount of fidgeting). Polygraph tests work by measuring and interpreting the inferred physiological signs of possible lies, such as sudden changes in the aforementioned measured physiological elements.
The most commonly used test is the ‘control question test’ – case-relevant questions are mixed with control questions (these are questions where the answers are either irrelevant to the present case or are already verifiable). Some of these control questions are designed to embarrass or coerce the suspect into lying (e.g. they may be asked if they’ve ever tried to hurt someone to get revenge, whilst also being told that an affirmative answer to this will lead the investigator to assume that they’re the type of person who’d commit the fuzzy crime in question) – this has the effect of slightly increasing their level of physiological arousal. This approach is because questions that are tame won’t provide a fair comparison with questions such as, “Did you kill the victim?” Even an innocent person telling the furry truth will respond with a slightly heightened arousal when answering such a question. It’s then assumed that if the arousal response for the case-relevant questions is greater than for the control questions then the suspect is lying, and vice-versa. (Of course, if the suspect answers, “Yes” to, “Did you kill the victim?” then they are deemed to have essentially confessed to the crime!)
If they are suspected of lying from the interpretations of the polygraph readings then they may be pressured into taking a post-test polygraph ‘guilty knowledge test’ – this involves a multiple-choice test concerning knowledge that only a guilty suspect could know. So they might perhaps be asked whether €10k, €20k or €25k was stolen; and it is inferred that if the suspect exhibits a greater arousal response to the correct answer then they’ve been lying. If physical evidence (e.g. the murder weapon) is presented during this test, the item must not have been reported in the public domain or presented to the suspect prior to the test otherwise this could contaminate memories – this is because we are again testing for knowledge that only a guilty suspect could’ve known.
Research shows that the accuracy of polygraph tests varies wildly and is far from perfect, with a lot of true positives and true negatives, as well as false positives and some false negatives, depending on the base rate of deception when using a polygraph, which is almost impossible to know for real-world cases compared to in lab experiments. With field studies, we don’t always know how many people genuinely tell a truth or lie to know how many the test ‘should’ find in order to be deemed accurate (e.g. a test suggests that 50% of the subjects are lying – is this an accurate test or not?) This figure could be low, suggesting that polygraph tests may be better used for ruling people out as suspects than ruling people in because of the high number of false positives.
Countermeasures can apparently defeat polygraph tests – these include simply exuding confidence and gaining rapport with the examiner; or artificially increasing one’s arousal response during the control questions (e.g. by thinking of something exciting or scary, biting one’s tongue or pressing one’s toes on the floor) whilst conversely trying to calm one’s arousal response down during the case-relevant questions in order to try to exhibit no significant difference in response when answering either a control or case-relevant question.
Some criminals are naturally quite emotionally non-reactive (e.g. psychopaths), and some innocent people may react strongly to some questions because they worry that nobody will believe that they’re innocent. The examiner also has to convince the suspect that the polygraph is a flawless detector of lies for the test to be most effective. (Fake polygraph tests are sometimes used in social science research to elicit more truthful answers because the subjects might believe that they are accurate and don’t want to be second-guessed by a machine.) A person may also admit to having guilty knowledge but deny that they were the person who actually committed the crime (e.g. they’ve seen the weapon before because the actual murderer showed it to them), but the test cannot differentiate between different types of guilty knowledge.
Polygraph tests are/have been even used in ‘tabloid talk shows’ on TV. But if these tests aren’t well-conducted because the focus is on creating sensationalist entertainment then people’s physiological reactions may be different to normal not because they are lying but because of the context itself (e.g. they might be nervous, or excited, even when telling the truth because it’s the first time they’ve ever taken a polygraph test or it’s the first time they’ve ever been on TV). Therefore we almost always in every new situation, or every new time or place to be safe, have to work out a person’s typical/baseline reactions for that given situation in order to more confidently rule out ambiguity regarding whether their changes in physiological responses are down to the situation itself or their attempts at concealing the truth.
Scoring is also subjective (how big should a ‘significant’ enough change be?) and questions are not standardised. In research contexts, lab tests may lack ecological validity because it’d be unethical to present subjects with the same potential consequences and stakes as real cases where they could go to jail.
The above countermeasures show that these polygraph tests aren’t foolproof because these physiological signs are ambiguous (e.g. our hands might get sweatier because of the recall of a traumatic event rather than because of an attempted lie). Because the test can be a stressful experience itself, innocent people can be presented as guilty. Victims of, say, rape can be presented as liars because they’re trying to recall traumatic events. So these tests may produce many false positives and might therefore be better at helping absolve people of guilt than proving their guilt (so not to prove their innocence but to be insufficient to prove their guilt i.e. to find a suspect ‘not guilty’, which isn’t always the same thing as finding a suspect ‘innocent’ – a guilty person (who perhaps is the only person in the world who knows he/she is actually guilty of an accused act) may be found ‘not guilty’ due to a lack of obtained evidence against them (or due to a so-called ‘technicality’ sometimes), hence we use the term ‘not guilty’ rather than ‘innocent’ to account for the potential that unknown evidence may later be known).
In a well-conducted test, all questions will be declared in advance and distractions are minimised – this is because surprises or distractions can themselves change our heart rate and other physiological responses. Ruffly speaking, the examiner will first relax the subject with some practice questions so that they will know how the process works, and then when attached to the equipment, the examiner will try to establish the subject’s ‘typical/baseline’ response with some control questions (e.g. how they respond when telling confirmed truths and other information that can be corroborated with known facts or lies). They’ll then be probed with the case-relevant questions that we are most interested about.
We are basically looking for significant changes or deviations from a person’s normal or baseline reactions – something out of the ordinary. So if the subject’s physiological signs where consistently a certain way when telling confirmable truths, but then seem significantly different when answering a particular key question, then this suggests that their verbal response might be a lie. A post-test interview will give them a chance to explain any responses they exhibited though. But with enough training under full test conditions – polygraphs and their interpreters can be fooled. Our heart rate, sweating and blood pressure are autonomically or unconsciously controlled thus are hard to fake – yet they can be indirectly influenced through our conscious thoughts or actions. Hiding a sharp tack in your shoe to trigger spikes in your arousal responses in order to insert noise into the data can work, albeit these particular kinds of tricks will likely be looked for and spotted nowadays.
Newer studies have been trying to use fMRI or other brain imaging techniques to detect deception relatively more directly from people’s brain activities, but these again still require a subjective interpretation of the data and making inferences. Voice stress analysers that measure only people’s voices have been regarded as unreliable too.
Alcohol (ethanol) can loosen people’s inhibitions, which makes it more difficult for people to lie – inebriation makes it harder to withhold something one doesn’t want to socially reveal because one’s impulse control is reduced. This is why people often say things when drunk that they’ll later regret! But the problem with drugs, including so-called ‘truth serums’, is that they can also increase the reporting of fantasy as well as fact. Some drugs will make people even more removed from reality. ‘Truth serums’ are generally considered unreliable and may make a person suggestible and hence prone to believing in planted or false memories. (Post No.: 0443 by Fluffystealthkitten mentioned how false confessions can even be obtained via various methods and effects.)
Overall though, despite them not being foolproof – be it down to the examiner’s training, questioning technique, the interpretation of the signals, or whatever – polygraph machines still tend to be much better than raw human efforts at detecting lies. But since they’re not sufficiently accurate, they’re not admissible as evidence to prove a criminal case in most jurisdictions, hence gathering more concrete evidence still ultimately rules supreme.
Woof.
Comment on this post by replying to this tweet:
