Are Liars Super-Duper Strategic?
Troubling recent findings call the "verifiability approach" into question.
Posted Feb 21, 2020
One of the hottest new things in applied lie detection is the so-called “verifiability approach.” The idea is that liars are strategic and actively avoid detection. Consequently, liars (compared to people telling the truth) avoid details that can be fact-checked. Because liars are predicted to provide more unverifiable information than honest speakers, counting the number of unverifiable statements is a good way to detect lies. Lots of unverifiable content signals a lie. Lots of things that can be fact-checked means the person is probably honest. This is the conjecture anyway. Do you buy it?
Anyone who follows the news closely might be skeptical. Donald John Trump says lots of verifiably false things. When a fact-checker calls him out, he often just calls this fake news (another verifiably false statement). But he is just one person, and likely not representative of the population.
With this as background, a recent experiment in Psychology, Crime & Law caught my eye. Liars and truth-tellers were compared in terms of the number of verifiable and unverifiable details in their statements. There was no difference in verifiable details. Liars did indeed offer multiple verifiable details. The bottom of the confidence interval was about 4 verifiable details per liar. The shocking results, however, were for unverifiable details. Honest statements contained, on average, just over 19. Lies, in contrast, contained just under 8. Opposite to the predictions at a rate of more than 2 to 1, truth-tellers in this data set provided many more unverifiable statements than liars.
It should go without saying that large data reversals are problematic for applied classification systems. This means that they can systematically get things wrong. Anything that can be significantly worse than doing nothing seems like a really bad idea. This was no mere failure to replicate. Being worse than no difference with an effect size of d = -.87 is an astonishingly large repudiation.
The verifiability approach exemplifies a concern I have with several approaches to deception detection. I think many theorists and researchers assume way too much meta-communicative competence. Liars are presumed to be highly strategic, thinking in advance, engaging in moves and counter-moves. Liars are understood as playing high-level chess.
I am very skeptical that most people are especially strategic or highly mindful communicators. Flying on autopilot and taking the path of least resistance are probably much more typical, both in deception and in other aspects of our communicative lives.
When I was a graduate student, I was interested in compliance-gaining. Researchers came up with a large list of different of strategies. But, if you watch what people do in their everyday lives when they want something from someone else, they just ask. Or, they just ask and give a simple reason. People can be strategic, but most of time they are probably not. It is a mistake to think otherwise.
Here is my advice. Do you want to know if someone is lying or telling the truth? If they provide a verifiable detail, fact-check it out. If it checks out, it is probably true. If not, it is wise to have some doubts. What matters is not if a detail is, in principle, verifiable but rather if it actually does hold up. Trying to get into liar’s heads presuming (a) one size fits all, and (b) high meta-communicative sophistication is an approach based on dubious premises.
Jupe, L. M., Vrij, A., Leal, S., & Nahari, G. (2019). Fading lies: applying the verifiability approach after a period of delay. Psychology, Crime & Law, DOI: 10.1080/1068316X.2019.1669594