Do you think you can detect liars? If so, according to a Nature article (https://doi.org/10.1038/s41562-023-01556-2 ) you are most probably mistaken. A meta-analytic test involving 24,483 people who had to attempt to pick truth/lie, found truth/lie discrimination no better than 4% better than random guessing. Not exactly a stunning achievement. You will read about cues such as liars avoid eye contact, and basically they do not. Liars have heard about that! Worse, truthful people do it as well. If you rely on behavioural cues, you find that there is a 96% overlap of these behavioural variables between truth-tellers and liars.
A current approach is to combine many cues, and the article reports that airport security personnel were trained to handle 92 cues. One problem is most of such a large number will be weak cues, but more significantly, who can do a rapid analysis of 98 variables in their head? Back to guessing, except those involved will have a very fancy name for the procedure that leads to guessing.
The problem is we need to end up with a binary judgement. This happens in many cases. A jury must decide on guilt or innocence, in a job interview someone must hire or reject. The action is discrete; twenty-seven ifs and buts have to be rejected, and your decision should be better than the toss of a coin. So how are such decisions reached? The usual way of dealing with too much information is to ignore most of it. The approach is to select a very few cues. If we want to do that for lie detection, we need to know the best possible cues.
In some experimental tests, when subjects could use any cue they liked, the accuracy was about 50%, which is what they would get from random guessing. However, when the same subjects were asked to make their decision based on richness of detail, success rose to about 66%, in other words two/thirds were correct. A second cue was whether the statement could in principle be verified; accuracy then rose the about 70%, but note that no verification actually took place. The key was, could it be verified? This was not actually done because in the tests decisions had to be made more or less on the spot, but it would be possible to ask what would be seen if someone went to inspect the situation to verify it or not.
An experiment was done where participants were asked to determine on a single cue (detailedness) or use multiple cues (detailedness, affect, unexpected complications, admissions of lack of memory). The success rate with a single cue was 59%, while with multiple cues, 54%. More information in the decision-making process led to worse decisions. Of course, you might well think this difference is not very great, and worse there may have been something in the material that biased the results, so just maybe the conclusion is only marginally significant. Another point might be that some of the additional cues were not very relevant. A memory lapse might be an excuse for not having tho0ught up the lie fully, but it can also be genuine. Who recalls all the fine details of having seen something but there was no reason at the time to think it was particularly important?
There is another problem with this sort of study, which was acknowledged by the authors. Subjects were instructed to lie or tell the truth. This may have led to deliberate “false cues”, and even if it did not, there is nothing at stake for the liar. Knowing there is a serious price to pay if caught out, liars may be more prepared to embellish their lies, and give additional details. Second, the success of this approach was demonstrated only where statements were about episodic memory and truth-tellers were both willing and able to provide specific details. In real situations, there may not be many details remembered when telling the truth, while a liar may produce more.
So, to summarize, those TV shows where there is someone who can pick a lie with infallible accuracy are, well, telling the audience porkies.