Schneier on Security
A blog covering security and security technology.
« Friday Squid Blogging: Dissecting a Giant Squid |
| The Two Classes of Airport Contraband »
September 22, 2008
India Using Brain Scans to Prove Guilt in Court
This seems like a whole lot of pseudo-science:
The technologies, generally regarded as promising but unproved, have yet to be widely accepted as evidence -- except in India, where in recent years judges have begun to admit brain scans. But it was only in June, in a murder case in Pune, in Maharashtra State, that a judge explicitly cited a scan as proof that the suspect's brain held "experiential knowledge" about the crime that only the killer could possess, sentencing her to life in prison.
This latest Indian attempt at getting past criminals -- defenses begins with an electroencephalogram, or EEG, in which electrodes are placed on the head to measure electrical waves. The suspect sits in silence, eyes shut. An investigator reads aloud details of the crime -- as prosecutors see it -- and the resulting brain images are processed using software built in Bangalore.
The software tries to detect whether, when the crime's details are recited, the brain lights up in specific regions -- the areas that, according to the technology's inventors, show measurable changes when experiences are relived, their smells and sounds summoned back to consciousness. The inventors of the technology claim the system can distinguish between people's memories of events they witnessed and between deeds they committed.
EDITED TO ADD (10/13): An expert committee said it is unscientific, but their findings weren't accepted.
Posted on September 22, 2008 at 6:10 AM
• 60 Comments
To receive these entries once a month by e-mail, sign up for the Crypto-Gram Newsletter.
It is probably significant that the defendent was a woman. India still has significant prejudices and inequities against women. If the defendent had been a man with some means, it may have turned out differently.
I have worked with EEG, and I agree that this is almost certainly a lot of crap. As far as I know, the work on what happens when you relive memories has been done with rats and epileptics, with electrodes implanted inside the brain, recording either from individual units or from small groups. I cannot imagine how you'd do anything similar by recording the massed signals you see on the scalp -- your data aren't even representing the same kind of electrical signal. And I also am not at all convinced it would work for memories of this nature. And wasn't the work in epileptic humans just published earlier this month in Science? Two weeks to market would be some impressive tech transfer.
I mean, I am not an expert, but given what I do know, this doesn't pass the smell test.
Even if this worked, it could be averted by blanking your mind and not listening to your crime being described. So it would create false negatives.
The scariest part is the level of false positives it must generate. Even if you did not commit the crime, there's no reason why reciting details of the crime might not evoke memories of other deeds they partook in.
This all sounds disprovable and testable though.
If this catches the terrorists, it's all worth it.
Assuming for the sake of argument that this technology works...
Would its use be more comparable to fingerprinting (i.e. acceptable in modern legal systems) or forced self-incrimination (i.e. illegal except in Gitmo)?
It doesn't matter whether or not this works, within a few years they're going to be requiring it to let you get on an airplane.
A.C. Grayling writes about this in his last New Scientist column. He condemns it pretty soundly.
Don't you mean, "if catches just one terrorists, it's worth it."
Actually, I heard the new version of this plan involves Vulcans...
This is "science" like the FBI claims to match a bullet to a box of them that were cast along with 1.5 million more of them from a huge ingot of lead. Its typical of "law enforcement" science which starts with a patsy and pins anything at all on that patsy. Sometimes when they frame the guilty, it blows back on them like the OJ case. After two days interogation by the brutal dallas police, Lee Oswald still maintained in public "I'm a patsy" a real political assasin would say, "I did it, I'd do it again" But powerful forces needed a patsy.
Seem's a bit akin to the old "braying donkey" lie detector. Perhaps if one is only willing to submit to the brain scan they are innocent...those who refuse have something to hide.
There are signals (p-whatever number of milliseconds) that can say whether a particular visual image is familiar to a person (although it's not clear to what degree of resolution or certainty with similar images). But going from that to a verbal description, especially when the defendant has likely heard a verbal description of elements of the crime repeatedly during interrogation...
This stuff makes polygraphs look good.
> those who refuse have something to hide.
Or, here in the US, they're exercising their right to not testify against themselves.
Urr. The polygraph is based on a similarly flawed assumption that increased heart rate/capilary perspiration/breathing rate can only be indicative of one thing - "creative thinking" (aka lying), and not of anything else (like, say, being nervous, remembering that you left the stove on, et cetera).
Granted, polygraph results are not accepted in our courts, but they are used in a staggering number of places, creating, arguably, a false sense of security - if the person passes the polygraph, they must be telling the truth...
Not all that far removed from the EEG mapping technique in question.
An expert committee said it is unscientific, but their findings weren't accepted: http://www.thehindu.com/2008/09/08/stories/...
"There is wrong diagnosis only in five out of every 100 persons subjected to the technique." so that's alright then
"The inventors of the technology claim the system can distinguish between people’s memories of events they witnessed and between deeds they committed."
Of course they do. They're not going to claim anything else, are they?
The fact that they feel the need to claim this implies that they can't prove it.
It's probably more "scientifically valid" than the TSA's "behavior detection officers" (who get a whole week of training). On the other hand, the BDOs once caught someone wearing a fake military jacket, so they have a proven track record of success!
@Greg: surely you mean "if it catches just one possible terrorist..."
Imagine using it for high crimes and misdemeanors. George W. would be immediately acquitted.
I have reasonable doubts about it myself. You can do math or you can use a calculator. The one may be faster, but getting to the result still requires thought. Results may vary because nothing is fool proof and there are lots of fools that are proof of this. Give them computers and they think they are wise.
Just think, if this technology plays out some day parents will finally be able to get a straight answer to the question, "What on earth were you thinking?"
@ Joel F:
"Or, here in the US, they're exercising their right to not testify against themselves."
Fair enough. Use it while you still have it.
This blog immediate conjured up thoughts of a certain "Simpsons" episode (Treehouse of Horror VIII: Easy-Bake Coven); Marge is accused of being a witch but is entitled to "due process". To which end she's thrown off a cliff with a broomstick and if she's a witch she'll be able to fly and return for punishment, if not then she's falls to an honourable death.
I'm guessing that the people who are innocent will generally submit to the supposed "brain scan" as they fear that if they don't they will forever be branded (and probably imprisoned) for the crime. Upon being scanned, the prosecution will then be able to prove these people's "guilt". Those that don't will leave the prosecution with less leverage, and a better chance of being proven "innocent"
Sounds like they've just copied the Scottish legal system and added some time-savings... erm... just kidding m'lord.
Remember, Lie Detector tests aren't admissable in US court, but just try to get a security clearance without taking one......
The first place to use such scans would be on death row. Release the innocent ones. Of course, not wanting to release anyone, they will refuse on the grounds that...it doesn't prove innocence?
JonathanW: That's great. We'll show that figure to the jury and explain that a positive result indicates a 95% probability of guilt.
Perhaps some of you gentlemen in the jury might think that 5% is still a reasonable doubt. If so, consider this: The 95% certainty of guilt is if we pulled a random citizen in for testing and - by pure serendipity - got a positive result. However, this accused was not chosen randomly; there was already a strong suspicion of her for such-and-such reasons. This cumulative evidence makes it quite a bit less probable than 5% that the accused is guilty. Essentially no doubt, certainly not a reasonable one, is left.
The prosecution rests.
On the vague assumption that the brain does show some reaction to somebody reading out the crime two things spring to mind.
1, How do the authorities know what happened to read it out (if they did then they would have sufficient proof against the person so it would not be necessary to go through the process).
2, Assuming what they are saying is just suposition, then how do they distinquish between the person reliving and a revulsion reaction?
ie if you said, "you stabed the victim in the stomach and they fell forward and you withdrew the knife as they died". This is a statment of possible facts that is not likley to induce either reliving or revulsion.
However if you said, "you thrust the knife with malice into your husband and felt the muscule fibers give and tear and watched joyfully as the blood spurted out in all directions and you felt intense pleasure at the way he folded over breathing his last painful breath as you drew the knife out knowing that he would never force himself on you again..." this would most likley evoke a strong emotional response of revulsion to what was being said which may give an even more powerfull response than the supposed reliving response.
Therfore as the test appears to be "show a response" then only a psycopath / sociopath is unlikly to show an emotional response.
Which tends to sugest passing the test is more likley a sign of potential guilt...
But as I initialy stated this a very lose assumption that they can actually detect anything with a scalp measurment. However if they used a Fast (N) MRI which can actually show brain activity and in which parts then we could have a high tech phrenology,
I saw a documentary recently that told the story of a blind man who can draw. His visual cortex lights up ... lots of activity.
When I read, I visualize the scene, the characters, the whole thing ... it is one of the reasons I love to read; I get to be the set director.
When I hear stories read aloud, I visualize what I am hearing. I wonder if a brain scan would show "experiential evidence" as I was visualizing the story I was hearing.
This report is chilling.
I'd love to see if we can get a bunch of false positives based on a large sampling, using some tests like these:
1. Read an account of a ficticious crime from third-person, record result.
2. Read an account of a ficticious crime from first-person, record result.
3. Read an account of a pleasant event from third-person, record result.
4. Read an account of a pleasant event from first-person, record result.
The accounts should be vivid, and the crime suitably graphic in details.
Now let this rest for a week or two. Repeat experiment. See if you get the same results. I think that people might start playing out scenarios in their heads, based on "what if I..." sorts of things. Like you might if you were accused of something, and starting to picture yourself in the role of the perp, whether it happened or not. Like you would if you were reading a novel with that eyewitness viewpoint. I think you might get an even stronger correlation if the initial accounts were in first-person. Also, based on what I have heard of false memory creation, you tend to get different results from unpleasant/upsetting things than from pleasant things.
Somehow, I don't think they've been doing those sorts of experiments with the technology... The general tendency is that if your tool is a hammer, everything tends to look like a nail.
"There is wrong diagnosis only in five out of every 100 persons subjected to the technique."
As shown by the 95% conviction rate, no doubt. QED!
I also see one of the inventors has applied for a patent on this, and for that reason no details can be published. Sigh...
What about people (estimated at about 4% of the population) that have some form of synaesthesia? I'd guess they light up the EEG and MRI, especially given those devices are being used to study the phenomenon in Aukland
"Lindsay Hearne, a psychology PhD student at The University of Auckland, is studying the brain patterns of people with synaesthesia, a condition where the senses cross. She is particularly interested in how the brain is wired in people who see colours for musical notes or letters of the alphabet.
Around 4% of people have some form of synaesthesia, ranging from the most common form of seeing colours for letters to the rarer form of spoken words producing taste. It is thought the condition may be hereditary, and may be caused by different regions of the brain being activated simultaneously."
"This seems like a whole lot of pseudo-science:"
I'm not at all surprised to hear you say that.
You clearly have the cranial features of a skeptic with something to hide.
What's your real agenda?
Dermot knows more about this than I do; you two should do lunch.
Wish I could get one of those for my pets. I'd love to know which one of them was digging in the yard. Just have to figure out how to recite details in a manner that will activate their brain in the right way...oh, does the method of story-telling affect outcomes?
Or perhaps even better, Nintendo could release the Wii-EEG for home use. Maybe they could call it an EEGii-board. Fun for the whole family. Never wonder again who ate the last cookie...
@Sparky: surely you mean "if it might give us a slim chance to catch just one possible terrorist..."
"...she's thrown off a cliff with a broomstick and if she's a witch she'll be able to fly and return for punishment, if not then she's falls to an honourable death."
and if she weighs the same as a duck...
Seems to be that this is conceptually no different than polygraph tests. Of course those aren't admissible in courts (in India either). Judges, especially in lower courts, do make mistakes like this. But the judicial system in India is robust enough - this conviction will surely be appealed and overturned in higher courts, if the EEG was the primary evidence.
Indian investigation agencies have been using those "truth drugs" (narco-analysis) frequently these days, especially in terrorism-related cases. The results aren't used in court, but they're invaluable in extracting leads out of suspects. If that results in the occasional wild-goose chase, that's a small price.
I am writing from india. engineering education, regular reader of this blog.
Many firends and I cringe in horror every-time they go for these scans of 'terrorists'. Many science people are protesting. No one is listening.
The police have a vested interest in carrying out these scans. They use it as a non-refutable excuse to frame the innocent. As a result, it is counted in 'solved cases' list of the police. Makes them look good, case solved, with minimum leg work.
Protesting and writing letters did not work. The science community tried and failed.
Anyone with a good idea on what Indians should do to stop this nonsence ?
"There is wrong diagnosis only in five out of every 100 persons subjected to the technique."
I wonder how this was assessed. Did they sample 100 persons whose brains were working feverishly because they faced a long sentence? And how were the false positives/negatives identified?
@Sparky: If it gives us a good chance of catching a terrorist, then its success rate is still inferior to simply shooting every suspect.
While it is true that polygraphs are not admissible in courts in the U.S. they still have a disproportionate affect on our criminal justice system since police and prosecutors generally take them seriously despite the pseudoscience and use them to guide prosecutions (if you are dumb enough to consent to one for police).
And, like psychics, they simply dismiss the well known number of false positives. For example, check out this story about a Grand Rapids woman who failed her polygraph and was prosecuted and then incontrovertible evidence of her innocence turned up: https://antipolygraph.org/blog/?p=61
It is strange that whereas so many professions are evidence-based these days, policing is still largely superstition and instinct. You'd think all of the overturned convictions from the Innocence Project, etc., would make police and prosecutors step back and take a look at how and why errors get introduced in the judicial system and what steps could be taken to reduce such errors.
Instead, no one seems to care and police and prosecutors seem to have very functionalist views of what they do.
I'd love to give this to some of James Randi's people. It's easy for researchers to construct protocols where something appears to work, because that's their job.
Perhaps it would be interesting to see the scan results for the investigators, who are by the nature of their work intimately familiar with whatever happened.
"Or perhaps even better, Nintendo could release the Wii-EEG for home use."
ok, I had to wipe coffee off the monitor from that one. :)
I figure the Wii-EG games are not far behind, featuring such classic titles as "Terrorist Hunter" and "Find a Pedophile" after all, it's about protecting the children, and how better to do that than have the Wii brain scan and profile people who play games with children?
Just set the Wii console and Hat-Controller to auto-scan and auto-submit on activation, then play away, knowing your children are safe. Who knows you could catch a wanted terrorist, pedophile, drug dealer, or dissident without even knowing you did it.
Evidence gathering cameras and microphones sold seperately.
If they are accused of a crime, then they are certainly guilty of that crime, otherwise, why would they have been accused? I say anyone accused of a crime should be lobotomized to extract the truth.
I used to work at the Triple Door in Seattle (a music venue). They had a special engagement one day where a whole bunch of wealthy looking people watched a demonstration video of this technology up on the big screen. It was about 3 years ago, I think. I got to hover around the back and watch too. It was pretty wild. I don't trust this kind of technology though.
'Better' tech for truth telling just means it's more difficult to convince people that it's easy to falsify.
@ Joel F:
"Or, here in the US, they're exercising their right to not testify against themselves."
Um,what part of waterboarding do you not understand?
That doesn't invalidate Joel's comment at all: You've the right to stay silent under waterboarding.
I am an academic researcher who works in the specific field of "mind reading" neuroscientific research, so I have a lot of experience trying to decode brain activity, and this "BEOS" concept is complete nonsense. To propose that it is possible to determine from EEG, which provides very poor localization of where brain signals are coming from in the first place, that a given experience was committed versus witnessed is fundamentally wrong on so many levels I don't know where to start.
Suppose you just want to evaluate this claim at a practical level, not to question whether or not our current understanding of memory would predict that this is even possible. How would you go about proving it? Brain mapping in particular requires delicately constructed experiments -- in my field, we joke that you might as well not scan patients in the first two versions of the experiment, but simply assume it didn't work and try to figure out why your design wasn't able to tease out any effect between conditions. In order to show that your analysis can actually detect the difference between witnessing and committing an act, you'd need to control for all other possible confounding factors: the particular act itself, the setting, the persons involved, etc. literally anything the subject could be thinking of that a powerful machine learning algorithm could pick up on. It is virtually impossible to do that in the carefully controlled setting that such experiments require. This is why any respectable neuroscientist will tell you that imaging based lie detection is impossible, at least until we fundamentally change the way all neuroimaging research is conducted.
So, yes, this is definitely pseudo-science, but more accurately it is a traditional snake-oil scheme to con the government out of money. It's really just psychics with electrodes. Unfortunately this time it puts innocent people in jail.
I, for one, welcome the literal manifestation of thoughtcrime.
"Anyone with a good idea on what Indians should do to stop this nonsence ?"
You could take up public protest, but unfortunatly in these draconian times it would be self defeating as your Government would immediatly say you are a "terrorist"...
It amazes me what Governments are getting away with these days. I fully expect to see the modern version of the "Spanish Inquasition", "Witch Trials", and "Burning at the Stake" to come into public expectation as entertainment.
After all watching some poor "happless hag" scream herself to death as she is slowly consumed by flames is much more interesting than watching some soap opera or a game of football...
We would however have to have teams of "witch finders" with their own Captins or "Witch Finder General" and a leauge table etc.
But as we don't belive in witches any more terrorists will have to stand in to be demonized...
I would love to see this tested on William Kristol, Richard Perle, David Frum, B. Netanyahu for crimes of their sort.
It would be intriguing to learn if this opaque science can detect differences between political and criminal thought.
That would be it's real value.
"the system can distinguish between people’s memories of events they witnessed and between deeds they committed."
What about memories of events in which they were the victim?
Even if this thing worked how would they tell from a brain scan whether my brain was lighting up because I was recalling that specific event or a similar event in which I was the victim.
What about if you have seen the crime that is being described in a movie and are able to remember it's details? Your brain will most certainly light-up but you would not have had commited the crime.
NotEnough: I presume that you were lobotomized. Too much TV perhaps? Don't you want some little waterboarding?
David: Being you, I'd quickly find another job. Imagine that you succeed and arrive to usable results. Can you imagine what will happen? The thousands ways such thing could be abused by the worst scum of the society sitting at the key post of decision making, in the army, intelligency... I'd not work for this if I was payed my weigh in gold every month.
Even if this tech is snake-oil today, it'll really work tomorrow. It's urgent to pass strong legislation (even a constitutional amendment, if possible) banning it RIGHT NOW before it gets into the hands of our Keystone Gestapo in the next installment of the Patriot Act.
Ok, let's forget my bias on this subject and look at the 5% of false positives.
5%...it doesn't sound like much but let's see what that means when used against population numbers.
It's estimated that in India as of July 1st 2007 the population was 1,129,866,154
5% of this approximately = 56,493,308
In the U.S.A the estimated population, according to the U.S. census bureau as of today, is 305,759,152
5% of this approximately = 15,287,958
Is that number of false positives acceptable??
Keep in mind that one of those false positives could be you.
Schneier.com is a personal website. Opinions expressed are not necessarily those of BT.