Schneier on Security
A blog covering security and security technology.
« Biliteral Ciphers |
| Federated Authentication »
March 28, 2011
Posted on March 28, 2011 at 1:10 PM
• 60 Comments
To receive these entries once a month by e-mail, sign up for the Crypto-Gram Newsletter.
48% of all statistics are made up on the spot.
Have they done any scientific work on why some of these tells exist? In particular, I'm curious about why the eye movements (supposedly) give you away.
Sounds like a bunch of BS to me. The sources include news sites and Paul Ekman, whose results are somewhat questionable, since (as far as I remember) he refuses to publish any papers and so nobody is able to reproduce his results, which makes me wonder whether they're there to begin with.
Okay, time for a lesson in Bayesian inference.
Consider the following two statements:
1) Probability that he is looking me in the eye, given that he is lying.
2) Probability that he is lying, given that he is looking me in the eye.
The problem with these sorts of studies is that they are used to infer the opposite of what the studies actually measure. Often, the opposite inference is unreliable.
1) Probability that he is Muslim, given that he is a terrorist.
2) Probability that he is a terrorist, given that he is Muslim.
The 1st statement has a high probability, the reversal is rather low. This has not stopped the reversal from causing much grief, aggravation, and wasted effort over the last 10 years.
The fact that anyone shows the "signs of lying" *cannot* be used to predict anything, even though the signs are backed by studies and correlate well with lying.
It's just an interesting correlation - a "hey Martha" moment.
@Okian Warrior: Yay for Bayesian inference! Didn't even think to apply it to this topic. I need to teach myself to use it at all times.
Straight out of the gumshoe handbook. Hasn't this stuff been discredited by actual science enough?
I've found that Bayesian inference is "turned around" quite a bit as a way to shape politics and social policy.
At the risk of being self promoting, here's my explanation of Bayes theorem. At the end is a page which shows 3 examples of how this is used to manipulate public sentiment:
We should search Muslims more often than non-Muslims:
o Probability that someone is a muslim given that they are a terrorist
o Probability that someone is a terrorist given that they are muslim
We must ban guns to reduce crime
o Probability that someone has a gun, given that they are a criminal
o Probability that someone is a criminal, given that they own a gun
Pot leads to hard drug use
o Probability that hard drug users have smoked pot
o Probability that pot users will use hard drugs
And so on. Learn to recognize the fallacy - it fairly permeates the political landscape.
Yikes! At the top of the page to which Bruce links, a banner ad from www.cia.gov urges me to consider "National Clandestine Service Careers"!
According to lab tests, that infographic is on average 51% accurate in detecting lies, peaking at 52% when used against 12 years old.
I call junk science. Particularly on the eyes and expressions, which can be heavily influenced by cultural background, the dynamics of the particular situation, and lots more. "I knew he was lying because he wouldn't look me in the eye" is probably the most common wrong statement that comes out of inexperienced interrogators.
By far the most inappropriate use of Flash I've seen all day. Flash adds nothing here except a broken scroll bar.
In addition, the site that came from (forensicpsychology.net) looks dubious as an expert site to me.
I can't say I find the list of sources at the end very compelling.
Given that some forensic folk believe in lie detectors, it doesn't surprise me one bit.
"1) Probability that he is Muslim, given that he is a terrorist. [...]
The 1st statement has a high probability"
A common myth, but a myth nonetheless, at least in the USA. Muslim terrorists get more press, but once you start doing research, you'll find that "domestic" terrorism is a much bigger problem than is commonly thought.
Where's the data? Show me the data ib wgucg these claims are based. Otherwise, I will assume the author of this graphic is a liar.
That whole "eye direction" thing was big hoopla, when, 5-10 years ago?
For fans of Paul Ekman, there is also the Fox show "Lie to me" with Tim Roth and Kelli Williams. Fun entertainment, especially the occasional snapshot display of politicians or other public figures doing some facial micro-expression prototype.
Hrm... sounds to me like lying is a lot like telling the truth. Funny how that works!
not ontopic story suggestion:
Spooks' secret TEMPEST-busting tech reinvented by US student
Young boffin blows gaff on mystery BAE submarine kit
www (dot) theregister.co.uk/2011/03/10/through_metal_comms_n_power_reinvented/
By Lewis Page
A mysterious secret technology, apparently in use by the British intelligence services in an undisclosed role, has been reinvented by a graduate student in America. Full details of the working principles are now available.
@ Okian Warrior:
No need for Bayes-watch here. What you cite are examples of the fallacies known for millennia: "cum hoc ergo propter hoc", and "post hoc ergo propter hoc". Both are pandemic throughout the worlds of science, "news", and popular culture. Too bad Logic 101 isn't required in high school (not referring to computer logic, of course).
@ ALL: All "lie-detecting" methods fail when needed most: against a sociopath who has no conscience, and therefore no change in thinking or physiology or movement when lying vs. truthing.
You don't even have to be a sociopath. A small dose of valium or xanax will already positively influence your chances of passing a polygraph or other test.
@ Dirk Praet:
Good point, and have also seen (in fiction) the opposite: Suspect plants tacks in his shoe, and rubs foot on tack when speaking the truth, thus throwing off the baseline by numerous false positives. Not sure how well this works -- does anyone know?
It has been tried and it doesn't work.
I thought it was a known fact that polygraphs are pseudoscience as well? Am I wrong? Yet, cops, Feds, courts and corps all use them.
I don't doubt that some of this stuff has some scientific basis or at least a basis in probability.
Where I'm not convinced is this "micro-expressions" stuff, especially when it's used to convince some people that other people aren't performing well. For instance, when the TV show Fringe came on, some critics said that audiences in focus testing didn't like Anna Torv (who plays lead character FBI agent Olivia Dunham) because she had "bad micro-expressions" and was thus a lousy actress and therefore the show was bad.
Where I call BS is the notion that everyone can and does detect such microexpressions and that it affects their reactions to other people in sufficient strength and consistency to be a real issue in such testing. Especially when according to the research, most people are not even able to detect such microexpressions in themselves or others.
It all smacks of overly broad application of what may be phenomena which is either so simple as to be useless or so complicated and affected by other phenomena as to be useless.
There are companies out there specializing in these sorts of predictions and I suspect many of them are very much over the line in the direction of fraudulent.
It's bigger in relative terms, anyway.
Even if "probability X is Muslim, given that X is a terrorist" is (lets just say) 5%, thats still a whopping much bigger number than "probability X is a terrorist, given that X is a Muslim" which is extremely small (like 0.0001% or smaller). There are over a billion Muslims on this planet, and probably at most a few thousand terrorists.
Conspiracy! The Friday Squiddy bit showed up today, and generated some comedy, so I was just about to add a brilliantly funny bit, but when I hit "Post", THIS is what I got!
An error occurred
No such entry '3801'.
It's a conspiracy, I tell you! I had just exposed Bruce's Friday squid designs as steganography intended to further the designs of Al Qaeda - or maybe even Cthulu! - when the rug was pulled out from under me!
Bruce, you can never hide the truth now! It's in the Way-Back Machine! All those squid posts!
Aside from what boils down to an affirming the consequent fallacy (mentioned by Okian Warrior in the context of Bayesian inference,) there's another hallmark of pseudoscience in this graphic.
Check out the bit about emotion. The graphic claims someone is lying if they start showing emotion too late, continue having emotion too long, or stop showing emotion too soon. That literally covers any response!
How to spot a terrorist: they are either too tall, too short, or too medium. With this simple rule, you'll never fail to spot one!
It's still there! :-) I just linked to it from Google Reader again. But you can't see it from Bruce's main page and if you try to post to it...
An error occurred
No such entry '3801'.
@ JJay: Thanks.
(You mean I can't believe TV shows? Oh, no!)
I'd like to see you prove that particular bit of bigoted propaganda.
Oh, you can't? That's because it's bigoted crap.
Ah, now the Friday squid post has vanished for real!
That was blogging fun. Makes me wonder whether I really want to start up my own blog. :-)
@larry seltzer "48% of all statistics are made up on the spot."
14% of people know that.
Reaction of a forensic psychologist to this article:
"It's a veritable potpourri of theories and findings and speculations, of variable veracity and even the best are far from infallible."
There are several unstated assumptions about this material -- sanity and rational thought of the liar being foremost.
As lots of others have said (yes, late to the party again), this is interesting from a humour point but is (on the whole) nonsense.
Where there are snippets of "correct" statements, they apply to such a small sample of the population that it become meaningless.
E.g. Yes some people touch their nose when they lie but in the vast majority of cases, it just means an itchy nose.
@bob: "I'd like to see you prove that particular bit of bigoted propaganda [ that most terrorists are Muslim]".
The "fallacy of the reversed conditional" is when you say "A implies B, therefore B implies A". A recent paper gives a good example:
Probability that John is dead, given that he was hanged, and
Probability that John was hanged, given that he is dead
This and other examples demonstrate that the rule doesn't always hold - sometimes it's true and sometimes it's not.
This is quantified exactly by Bayes theorem, which gives a scaling factor to be used when reversing conditionals.
Sometimes the scale factor is close to 1, which means that reversing the conditional makes sense. Sometimes it's low, which means that reversing doesn't tell you anything.
I believe that most terrorists are Muslim because I read about a suicide bomber almost daily. Over time, this has racked up a score which I think dwarfs the contribution from any other sector.
The facts would seem to bear this out:
Going down the list, the vast majority of events are caused by followers of Islam.
Bob, I can recognize the correlation without using it to paint an otherwise peaceful religion in broad strokes.
I trust the math.
On the other side, you seem susceptible to this type of manipulation. Because I believe that most terrorists are Muslim, you think that I am bigoted.
Let reason and logic prevail.
Good to know, but this is better: Here's how the TSA detects witches:
"If you were the boss at a U.S. government agency and one of your employees complained that she was afraid of a co-worker's religious practices, what would you do?
Would it change your decision if the religion were Wicca, and the employee feared her co-worker because she thought she might cast a spell on her?
Here's how the Transportation Security Administration handled it:
It fired the witch."
"1) Probability that he is Muslim, given that he is a terrorist."
In the US, from 1980 to 2005, that probability is about 6%. cf:
and an analysis of those figures at:
Actually that's not quite what the figure is saying - it's more like "Probability that a given act of terrorism was motivated by Islamic or pan-Arabic extremism" is 6%. If a member of the Red Brigades, the Animal Liberation Front, or the Ejercito Popular Boricua Macheteros committed an act of terrorism, and also happened to be Muslim, then it gets counted under Communist-, Leftist-, or Latino-motivated terrorism (that latter being, as far as I can tell, code mainly for Puerto-Rican independence-motivated) - the religion of the perpetrator in that case is beside the point.
That said - by the looks of those figures, given that an act of terrorism has happened on US soil, the overwhelming likelihood is that the perpetrator is either Catholic or atheist (assuming stereotypes of Latinos and leftists hold in this case)
Penn and Teller demonstrated the tack approach vs. polygraphs on their show and alleged that it can work, but it looks like the sort of thing that requires practice and a degree of self-control not available to everyone.
It's also been demonstrated that polygraphs can be beaten by simply believing that they don't work. At least that's what I was taught in psyche 101 decades ago.
@DragonFrog: "In the US, from 1980 to 2005, that probability is about 6%"
You are cherry-picking a subset of the data in order to disprove a population trend.
Would you care to address the original statement? Meaning - what is the probability worldwide now?
("Now", meaning "extrapolated from past couple of years' trends".)
Since what a polygraph really measures is the degree to which you're physically amped up, anything that relaxes and reassures you certainly doesn't help the interrogator.
@ Petréa Mitchell and Shane:
I was wondering about the opposite approach of Shane's: Let's say that during the entire interview, I'm picturing horrible things. (The airplane I'm in just had a wing fall off. We're plunging to the ground. Nothing can be done about it. Here comes the impact...) That's going to produce such a strong emotional response, if I picture it vividly, that it seems it would obscure changes produced by lying - assuming that I keep that and other horrible things in mind during both truth and lies.
If the interrogator says anything, there's plausible deniability:
For a job interview: "I'm nervous, because I've heard that polygraphs are unreliable, and I'm afraid of losing this job because of a mistaken result." (could easily be true)
For interrogation, whether by LEO or by foreign or criminal element:
"Of course I'm nervous! I'm locked up here, with no rights, no way to get out, and I'm probably going to be executed no matter what I say!" ... etc.
What do you think?
"Would you care to address the original statement? Meaning - what is the probability worldwide now?
("Now", meaning "extrapolated from past couple of years' trends".)"
I wish I had some better statistics.
Here's something interesting to ponder though - compare the article you linked to this one: http://en.wikipedia.org/wiki/...
How many ETA terrorist actions do you find in your link? I find one. How many do you find in the article on ETA attacks? I count eleven, not counting minor arson.
So, is this a "population trend", is it under-reporting of terrorism committed by Basques, or is it something else?
That's to say nothing of acts that might be labeled terrorism, except that the acting organization is "not a terrorist group" (everyone probably has their favourite group for this category).
Also - why "extrapolated from past couple of years' trends"
You wouldn't be trying to cherry-pick a brief trend to disprove a long-term historical tendency would you?
Anyone who has been through the NSA polygraph test, as I have, will know it's not the machine -- it's the operator (not to mention the people behind the one way mirror).
I can fool a polygraph more or less at will -- I owned one for awhile and learned how. I didn't even try at NSA, for one thing I had a big interest in telling the truth.
The woman who examined me should be at the TSA, should she be willing to stoop that low.
Perhaps one of the best hinky-detectors alive at the time. Bruce would be proud.
The machine is just a prop, they know it, it works on some people. The investigator is the thing, and some of them are really good at their jobs. They can make the machine work better too -- by quick follow ups, surprises, what amounts to constant calibration of *you*. I was very impressed, and glad I wasn't trying to lie.
@dragonfrog, I think we're sufficiently far afield of the topic that we should stop.
Please bear in mind that my point is that *even if* most terrorists are Muslim, that says nothing about Muslims in general because you can't reverse the conditional.
I'm trying to reassure people about Muslims by pointing out the fallacies that people [politicians, news media] use to drum up fear.
That's the opposite of bigotry. I believe in the math, *and* I believe that making strong statements helps the problem, even if only a little.
Beyond that, the statement "most terrorists are Muslim" may be wrong, but the position stems from evidence and isn't completely devoid of merit. It's not a bigoted position.
We need the ability to state our views without automatically pressing the hot buttons and calling names. If I'm wrong, show the evidence and I'll change my opinion.
We're getting bogged down in semantics, in "what time period" and "what geographic area" and "what is a terrorist act".
Let's not lose sight of the original point.
@stvs Here's how the Transportation Security Administration handled it:
It fired the witch."
Unbelievable!! And without even weighing her against a duck!! What fools they are for not applying plain logic!
@ Doug Coulter:
Supporting what Bruce has said many times: that *well-trained* humans are much better than machines (cameras, computers, poly) at detecting hinkyness.
While *poorly-trained* humans, like TSA, are of dubious value, even using invasive technology and practices.
Thanks for sharing.
Getting bogged down is what makes the internet fun.
While I fully agree with your analysis and reasoning, I dont agree with one of your initial assumptions - that most terrorists are Islamic.
We are all cherry picking data unless we are going to account for every terrorist act since the dawn of humanity - in which case there certainly would be a lot more non-Muslims carrying out attacks than Muslims.
You have, to an extent, committed a fallacy of assuming that most attacks = most terrorists. This is not always true and would need a more detailed analysis of how many people have to be involved in a suicide vest at an Israeli checkpoint vs identifying an off-duty police man and putting a bomb under their vehicle.
Also, Wikipedia is far from a definite source of data for even the number of attacks carried out.
As an example, in Northern Ireland over the last 12 months there have been almost one "terrorist" act carried out each week. Most of these are the "trivial" events, such as pipe bombs being thrown at police stations, which result in little more than a local news headline. They are still terrorist attacks.
None of this even addresses the fundamental issue of what is a terrorist. If we include "freedom fighter" activity then the issue becomes a lot more muddled.
It is fairly safe to say, however that some truths exist, for example:
If a bomb explodes on UK soil the probability is close to 100% that it was detonated by a Catholic Republican.
However if you pick a random Catholic Republican, the chances of him being a terrorist are approaching 0%.
Most terrorist attacks in Northern Ireland are carried out by Republicans.
Not all Republicans are terrorists.
For some reason, we accept and understand this. But when the subject is changed to Muslim we (society on the whole) get all confused.
Avoiding eye contact.
Touching ear, hair and nose.
Today I learned that I lie even when I'm telling the truth.
Or even when I'm not talking at all.
Many of the traits/movements of liars, also apply to people on the Autistic Spectrum who aren't lairs.
I reposted the link to the chart to the asperger community on LJ. Below are several responses.
However, many Aspies exhibit many of the traits ascribed to liars in the chart below. Thus leading to more prejudice against Aspies.
This information has actually helped me spot lying in neurotypicals, and has helped me change the way I look so people don't think I'm lying, but I agree that it leads to more prejudice against Aspies. I find it upsetting that I have to change the way I look to begin with just so people won't think I'm a liar.
I feel like printing this out and stapling it to my wall...or my t-shirt...
They tend to believe they are experts at spotting liars, so assume Aspies are lying.
Jeez! They might as well have called it "How To Spot an Aspie." =P
Sure is a good thing no people with an interest in deceiving others will read that page and adjust their behavior according.
Rhialto, I agree. A very fishy looking page. Fortunately NoScript protects me against such funny business.
So, I never saw the infographic but I do not consider that as a big loss.
@doug. You have it exactly right. Microexpressions can be useful, but less in casual conversation. You have to jostle someone to react only to the question or what they are saying. People could be reacting to a thought (wondering mind). I would not want to think someone lied about a material matter when they are thinking of the stinkbomb diaper the little one dropped this morning, that the little one peed in moms face changing that diaper, or that their tooth hurts...You want a raw reaction that is focused. Other than that microexpressions work perfectly, every time. ;)
@GreenSquirrel: I was just about to post to make exactly that point.
Looking at the list of terrorist events from 2010 via the link posted earlier, I notice that most of these events take place in extremely volatile locations. Given the situation and the fact that these kinds of events are likely to happen under these conditions, it makes sense that the perpetrators of most of the acts of terrorism will, to some degree represent the ethnic and/or religious makeup of that location or conflict.
I don't see how it is safe to assume that the majority of other people within that demographic are similarly predisposed to committing such actions. Following that reasoning nearly every Irish citizen would be on a no-fly list given that we've had more than our fair share of atrocities committed in our name.
I had a quick read through the infographic and found it to contain very little of real substance. Note that every single one of the points prefaced with, "A liar will" has the word "often" following in smaller type. This is a classic example of someone manipulating the presentation of data to reinforce their own agenda and so accusations of a similar nature could apply equally well to the authors of this amusing but inconsequential graphic. They might as well have added "DVD Box sets for series 1 to 2 of 'Lie to me'" to their list of references which would have had more credibility than some of their other sources.
After looking at the "Forensic Psychology" site it looks very much to me like the work of people who feel that they are experts in a field of interest but have no real working knowledge of their chosen subject.
@Kevin - well said.
Also, given the recent events in Omagh, it seems we have to chalk up another pointless murder to the hands of the non-Islamic terrorist.
To their credit, the people of Ireland (both bits) and the UK are all reacting in what (I think) is the correct way following a terrorist attack.
It is a shame that we all know if this had been an Islamic terrorist attack there would still be a dead person and a destroyed family, but the national reaction would be significantly different.
Hm... I have seen a study that shows pretty conclusively that eye movements are not correlated with cognitive processes. That doesn't mean everything else they say is also inaccurate, but it does cast doubt on the whole thing.
Schneier.com is a personal website. Opinions expressed are not necessarily those of Co3 Systems, Inc.