Lie Detection by Strategy Manipulation | Hacker News Reader

Lie Detection by Strategy Manipulation

sciencedirect.com

38 pointscwwc19 comments

> Ideally, interview protocols to improve lie-detection should (a) encourage forthcoming verbal strategies from truth tellers and (b) encourage withholding verbal strategies from liars.

I'm not sure this is always true.

I've been playing A LOT of card game named Cheat, it's basically training how to lie and how to detect lying better.

The game is about losing all your cards placing 1 or 3+ of them at once on the stack, but they have to match the previously announced card by suit or color (and then suit has to be adjacent - like 3 of spades matches 4 of spades but not 5 of spades). You place them backs-up and have to announce what they are (you only announce the part that matches - like "three aces") and you're allowed to cheat (putting more cards than you said or placing other cards than you said you did) but if someone checks you immediately after your turn - either he or you take all the cards from the stack depending on who was right.

There are lots of other variants but that was what we found the most interesting.

It's a very good simulator of lying because there is hidden information and people might be able to tell when you lie (if you announce three aces and somebody has two for example). But then people can also ask around if others have aces to check you. But other people might lie to make them take cards.

There's lots of strategies including pretending to lie and pretending to pretend to lie etc. It goes many levels deep. Simple strategies like "provide many details to show you don't lie" or "provide few details to pretend you lie" usually lose to more sophisticated ones.

Keep in mind that this is a game designed for fun not realism. There is no prize for tricking the police into arresting you.

Yeah that's the biggest difference. Among Us (or Mafia game on which it was based) is closer probably. But still there are smaller scenarios that can play out in Cheat where the situation is close (you have 2 players providing contradicting information and you try to guess who's right).

Anyway my intuition about detecting liers is - people who provide too much or too personal details without being asked about them are more suspicious than the default, not less.

There's a few proverbs in my language about that ("only the guilty explains himself" and "hit the table and the scissors will speak out") - the first one obviously stupid on the face value but I think it is supposed to mean the same thing - that the guilty will have more explanations ready faster because they thought about them more and focus more on being considered innocent.

Maybe that's why the "explain to them" part is important.

> I've been playing A LOT of card game named Cheat

A difference with this research would be whether it's credible to the liar that you already have, or can later access, unspoken information from an outside source.

Your goal as a liar is in general to construct a credible, consistent story.

In the controlled laboratory conditions, the liars can easily believe that the interviewers might have access to some unstated outside information, so constructing a story that remains consistent with that information is immediately more difficult. You've at least got to weigh up how you might incriminate yourself, before saying anything.

In Cheat, a liar knows you won't have access to any other outside information than what they can also see. (Unless you have two aces, in which case they are already incriminated and story telling will do no further harm.) That means their lengthy story-telling can, with sufficient experience of the game, remain safely consistent with the information you have.

> In the card game Cheat, your opponent knows that you don't have access to any more information than what is visible.

You can ask other people if they have any more aces to verify "3 aces" claim by one player.

But then they can lie of course :)

So you have access to other information, just not very trustworthy.

> So you have access to other information, just not very trustworthy.

This is equally visible to the liar; you don't have any information that they don't also know about and can control for.

[I already edited this in the original post for attempted clarity, apologies for any confusion]

Is it good or bad for a society to learn about lying? I would have thought bad until I’ve awoken to the idea that deception is a regular practice of those who do harm, and a regular weakness of those who have harm done to them, but it doesn’t work the other way. That is, the ability to deceive doesn’t make one deceptive. That comes from the desire to do harm, which often comes from a lack of ability or imagination to do well without harming others. Harm is always felt as a necessity. So intelligent people do not practice or learn about deception, and therefore are most vulnerable to it. Maybe they need to learn by practice, the way their attackers do. A society needs trust to function, to build slowly, but all it takes is a little bit of deception from a small group of people, to destroy everything very quickly.

It's like martial arts IMHO - it's fine (maybe even beneficial) as a hobby but if you go around beating up unwilling people it's obviously evil.

I see it as psychological chess.

It's good. There will always be those that deceive to cause harm. Knowing the tools of deception helps you know when you're encountering one of those types. Now that you have the tools, it's up to you whether you, personally, want to use them for good or evil.

I played a version of this in school and found it a bit simple (while still fun) after a while. My optimal (and often successful) strategy was to make plausible moves (announce cards I have but play something different) and play with as little emotional reaction as possible.

A truly challenging manipulation game, in my opinion, is Among Us, because you can freely question people and potentially have to come up with a plausible lie on the spot.

Among Us is basically Mafia and I played that too, but it required too many people and took too long to play it on the breaks.

It probably depends on the exact variant you played and on the people but I preferred Cheat. If you only did plausible moves you often lose to people playing more risky (but that changed with "meta").

''Essentially, the AIM method involves informing suspects of these facts. Specifically, interviewers make it clear to interviewers that if they provide longer, more detailed statements about the event of interest, then the investigator will be better able to detect if they are telling the truth or lying. For truth-tellers, this is good news. For liars, this is less good news. Indeed, research shows that when suspects are provided with these instructions, they behave differently depending on whether they are telling the truth or not. Truth-tellers typically seek to demonstrate their innocence and commonly provide more detailed information in response to such instructions. In contrast, liars wish to conceal their guilt. This means they are more likely to strategically withhold information in response to the AIM instructions. Their (totally correct) assumption here is that providing more information will make it easier for the investigator to detect their lie, so instead, they provide less information.'

......

This asymmetry in responses from liars and truth-tellers - from which the AIM technique derives its name - suggests two conclusions. When using the AIM instructions, if the investigator is presented with a potential suspect who is providing lots of detailed information, they are likely to be telling the truth. In contrast, if the potential suspect is lying then the investigator would typically be presented with shorter statements.''

''We found that when the AIM technique was used, it was easier for the interviewer to spot liars. In fact, lie-detection accuracy rates increased from 48% (no AIM) to 81% – with truth-tellers providing more information.''

Isn't this only going to be the case for liars who haven't read this research?

As someone who went to the US Army’s interrogation school over a decade ago. This checks out. I’d also add that a liar will also suck at creating a realistic and coherent timeline.

re:timeline, I've always heard that liars recall the story they're making up in order; truth tellers tell the most important and interesting bits first. I.E:

0) Sorry I'm late 1) I woke up 2) Got in the car 3) Traffic was really bad 4) I got in a car accident on the way here

vs.

0) Sorry I'm late 1) I got in a car accident on the way here 2) Traffic was really bad

Because liars are often making it up as they go, they tell stories chronologically, filling in the details as they go, unable to pre-recognize the interesting details.

Obviously, its not always that cut and dry, but its a hint.

This research reminded me of this scene from the film Uncle Buck https://www.youtube.com/watch?v=1CrPAl1aDzM

Buck: Did you brush your teeth?

Miles: Yeah. You can even feel my toothbrush.

Buck: You know, I have a friend who works at the crime lab at the police station. I could give him your toothbrush and he could run a test on it... to see if you actually brushed your teeth... or just ran your toothbrush under the faucet.

[Buck leaves, as Miles gawks]

Maizy: If that's true, we're gonna REALLY have to start brushing our teeth.

Research presented by the author here (no paywall):

https://theconversation.com/spotting-liars-is-hard-but-our-n...

You can also read the original posted article at sci-hub.se.

Reply on news.ycombinator.com