Thread by @drjuliashaw

Think you can catch a liar?

My recent article made the cover article for BBC @sciencefocus. It's about what 100 years of lie detection research has taught us.

I interviewed some of the world's biggest names in deception detection research. What they said surprised me 🧵

Dr Luke (@RabbitSnore) explained “The construct of deception is more complicated than a lot of people treat it”

We use ‘lie’ to refer to someone pretending an outfit looks good, a partner hiding an affair, or a murderer pretending to be innocent. But are these really comparable?

When it comes to lying, or deception as researchers like Dr Luke (@RabbitSnore) prefer to call it, “at this point we are not talking about the same thing. Even superficial things like the length and type of communication matter”.

“The empirical work on deception detection is massive,” explained Dr Pär-Anders Granhag, a deception research powerhouse. “But the only single issue that a large majority of the experts agree upon is that gaze aversion is not a diagnostic cue to deception.”

70% of the experts in a study by Dr Luke (@RabbitSnore) agreed that liars *don’t* act more nervous than truth tellers.

Prolific deception researcher Dr Vrij explained to me that the most prevalent misconception about deception is “the idea that nonverbal lie detection works”.

Results of the expert study suggest those trying to use non-verbal lie detection methods should proceed with *extreme caution*

Including those currently promoting tech solutions like the polygraph, video analysis for emotional leakage, brain fingerprints & changes in vocal pitch

So, are there any effective methods for spotting a liar?

According to the study by Dr Luke and colleagues, there is one cue that is promising; 72% of experts agreed that liars provide *fewer details* than people who tell the truth.

This is echoed by Dr Vrij, who argues that we shouldn’t look at how people behave, we should examine what they say.

He told me “several verbal indicators are diagnostic” including the total number of details and ‘complications’.

Dr Luke and Dr Granhag have proposed a "Shift-of-Strategy" approach to get info that suspects are concealing.

This is a strategic use of evidence which involves drip feeding what you already know to the suspected liar, without accusing them of lying directly.

Dr Granhag argues this works because, “If a person changes his or her story when you present parts of the background information that you hold, you are on your way catching a lie”.

“Most annoying is the assumption that comes from TV shows… that lead the general public including professionals to think they can catch an individual liar.”
Dr @AminaMemon1 told me, a world-leading researcher on lie detection and investigative interviewing.

Police following a hunch, based on the stereotypical profile of a liar, may use coercive tactics that lead innocent people to confess

Dr @AminaMemon1 advocates for a neutral, fact-finding, approach to interviewing that steers clear of trying to guess at whether someone is lying.

*The most interesting thing I learned*. There’s a bigger issue lurking behind the research on deception that Dr Luke (@RabbitSnore) and others have pointed to;
Maybe the reason we haven’t found universal cues to deception that experts agree on is because they don’t exist.

For the past hundred years deception researchers have almost exclusively taken a ‘nomothetic’ approach

They are looking for the ‘laws’ of deception, cues that apply to everyone

But perhaps this kind of one-size-fits-all approach hasn’t worked because everyone lies differently

To understand cues to deception, argues Dr Luke (@rabbitsnore), researchers need to adopt an ‘ideographic’ approach

Ideographic research focuses on what makes each of us unique. This would involve creating a personalised profile of how an individual lies

Example of ideographic approach:

@SophieVanDerZee developed the first deception model tailored to an individual. She used a fact-checked database of Tweets by a recent US president, & found the language he used when he lied was systematically different from his truthful tweets.

Once @SophieVanDerZee made a personalised profile, her team could predict whether his Tweets were untrue with a high accuracy rate of 74%.

This kind of personalised deception detection model can work for those who already have a large online presence where they lie a lot.

Ai can assist with collating & examining these existing data. But what about for those of us who are less present online, or who don’t lie in posts?

Some things you can fact-check, but most everyday posts and messages are so personal that it’s hard to even identify them as lies

“There is no guarantee that a machine learning model is going to actually work” Dr Luke told me.

It seems clear that a seismic shift in the science of lie detection is now underway. It’s time to move away from what Dr Luke (@RabbitSnore) calls “crude averages”.

“People are a little too fascinated by having a cool little trick to catch someone in a lie.” says Dr Luke.

Liars can evade detection partly because they also know the stereotypes and play into them.

And our confirmation bias can also make us overconfident – we disproportionately remember the times we caught liars, and don’t realise all the times we failed.

For the times we do succeed, Dr Luke isn’t convinced that the cues we *think* we used, worked.

“Think about the last time that you caught someone in a lie. How did you know? It’s probably not because someone looked up and to the left. You probably had some evidence".

Stereotypical cues might actually make you worse at catching a liar.

“It's better to trust your own detective work and check what people say against evidence” says Dr Luke. And if you can’t find evidence? “then proceed with caution and don’t be too confident”.

End of 🧵.

This 🧵 on 100 years of deception detection research might be interesting to @crest_research @portsmouthuni @EAPLstudent @UCLBrainScience @UCLLaws @JohnJayResearch @APLSsc @BPSOfficial @goteborgsuni @anneliesvrede

You might like (or share?) this thread that summarises current research on lie detection & dispels lots of myths @CivilLitTweet @BarristerSecret @AdamRutherford @bengoldacre @FryRsquared @joshbakerstory @CarolineSteel3 @LindaGeven @LinRod