Brain-Scan Lie Detection Rejected By Brooklyn Court 197
blair1q writes "A judge in Brooklyn has excluded Functional Magnetic Resonance Imaging (fMRI) lie-detector evidence from a trial there. However, the decision will not set a precedent, as it was made without even conducting a hearing on the method's validity, but on the principle, argued by the defense, that 'juries are supposed to decide the credibility of the witness, and fMRI lie detection, even if it could be proven completely accurate, infringes on that right.' That principle can be tested in later hearings, such as one scheduled for May 13, 2010, in Tennessee; in this case, the defense wants to use fMRI evidence it has already collected to prove its client is innocent. fMRI has been shown to be 76-90% accurate. That number seems significantly larger than the rate of false convictions."
Re:The statistical value of "resonable dobut" = 0. (Score:4, Insightful)
Nice tech, but it's not there yet.
What happens to the right to remain silent when it is there? The British have already gutted this right. Not that hard to envision the same happening here.....
Not Very Accurate (Score:5, Insightful)
fMRI has been shown to be 76-90% accurate
That's certainly better than a weatherman but not good enough to convict someone.
It is better than a jury of Bobs (Score:3, Insightful)
If it is better than what we have for false convictions than why prefer human prejudice/error over machine error. It seems to me one of those is far more likely to improve than the other and I'm not talking about Homo Sapien's ability to use critical thinking skills when confronted with conflicting emotive/subjective versions of events.
Also it is would only be a portion of the evidence which the case depends upon currently for it to reach a verdict. Anyone who compares this to lie detector machines does not understand that lie detectors are little better than a coin toss. This when properly used has been shown to top out at 90%.
Re:It is better than a jury of Bobs (Score:5, Insightful)
If it is better than what we have for false convictions than why prefer human prejudice/error over machine error.
Because one of the reasons we have a jury system is to provide a check and balance on the ability of the government to lock people up. Jury nullification may be a bad word in the modern legal system but it's still there.
Re:It's the legal industrial-complex, duh! (Score:3, Insightful)
But of course the whole point is that facts, taken in total, tend to point the way to the larger truth at issue. How else would you have us do it?
Re:Because they are unreliable. (Score:5, Insightful)
Absolutely no lie detector will ever be able to tell the difference between "this is true" and "at this moment, the suspect believes that this is true". With some mental training (or, you know, schizophrenia) it's entirely possible to temporarily convince yourself that the sky is purple, and there's basically no way any machine will be able to pick up on it in the foreseeable future.
No judgment should ever rely on "the machine says he thinks its true, therefore he is guilty" - no matter how accurate the machine is.
Re:Because they are unreliable. (Score:3, Insightful)
More importantly, lie detectors can only tell what the subject believes to be true. Given the number of people in America that believe that the world is flat, that Elvis is alive, that George Bush masterminded the 9/11 bombings, that Oswald didn't kill Kennedy, or that off-sea oil-rigs pose no risks, I think it is safe to say: "All sorts of people in America believe things that are not true."
This is a huge problem for witnesses at accidents. 5 different witnesses will give the cops 5 different stories, and then when the case gets to trial, 5 additional slightly different stories. People remember things in different ways, and some believe strange things. It is a fact of life.
Re:Because they are unreliable. (Score:4, Insightful)
Instead, we rely on "these 12 guys who have been struggling to stay awake during the proceedings think it's true, therefore he is guilty."
And he did very well in excluding it (Score:3, Insightful)
Re:The statistical value of "resonable dobut" = 0. (Score:3, Insightful)
What you say would be correct if people were being convicted only on the basis of this fMRI lie detector test. In practice, how you get to a 1/1000 error rate is by combining several less reliable sources. For example, convicting somebody on the basis of one witness is crazy, but convicting them on the basis of 10 witnesses is reasonable. (Given certain assertions of statistical independence etc).
Classic statistics problem (Score:5, Insightful)
This sort of thinking (that if the accuracy rate is improved enough it will become a valid way of determining someones guilt) shows a fundamental misunderstanding of statistics. It is the same reason blanket drug testing doesn't work and medical screening can sometimes be a bad thing.
Let's imagine for a moment that this lie detector technology has been perfected to a 99.99% accuracy rate. Since the test is so accurate, we decide that whenever a crime is committed, we will just have everyone in the area take the lie-detector test, asking them the question "Did you commit the crime?". Clearly, when someone fails the test, they are 99.99% likely to be the criminal. Right?
Except no. In cases like this (where the average person is much much much more likely to NOT be the criminal, the error rate will overwhelm the actual guilty-rate. If we are testing everybody in an area, then we can suppose that each person we check has an average chance of being the criminal of about 1 in however many we are checking. If this number we are checking is very large, then we are CERTAINLY going to have quite a few people who are found to be guilty on the test but are actually innocent. It will pick out more innocent people than guilty people.
While this sort of statistical phenomena will not take place if we don't giving blanket tests to everyone and limit the test to people who we already believe are very likely to have committed the crime, we as a society have a very bad tendency to not understand the statistics and think we should just give everyone the test and let the results tell us who is guilty. If you doubt this, just look at how many people think we should have a DNA database that everyone needs to join (so we can just run any DNA found at a crime scene against it). This combines the birthday paradox with the statistics I described above to create a situation where we have a very real fear of false convictions, exacerbated by the fact that people who are relying on this evidence (juries) do not realize that even a test with 99.999% accuracy can have a very high false positive rate in these sorts of circumstances.
Check out http://en.wikipedia.org/wiki/Bayes'_theorem#Example_1:_Drug_testing [wikipedia.org] for more info on the math behind this.
Re:The statistical value of "resonable dobut" = 0. (Score:5, Insightful)
Citation needed.
"Proof beyond a reasonable doubt" is not a formula or a slogan, to be sold, like Ivory soap, as "99 and 44/100% pure."
It only means, that in the light of all the evidence presented, the jury can in good conscience say that the defendant's guilt has been proven to their satisfaction and that any questions which remain will not alter their decision.
Re:Because they are unreliable. (Score:2, Insightful)
"these 12 guys who have been struggling to stay awake during the proceedings, and just want to go home, think it's true, therefore he is guilty."
FIFY
Re:Classic statistics problem (Score:3, Insightful)
While I agree with your post I can see the slippery slope the OP is talking about. Where does allowing something like this lead? If it is so good, shouldn't it be mandatory, etc?
Re:Lie Detection (Score:2, Insightful)
Another problem with "lie detectors," and a good reason that juries rarely ever hear about them, is that juries tend to give them undue deference. You can get a competent defense counsel to present evidence to a jury that they're not reliable, have a lot of false positives, etc etc....and at the end of the day, many jurors will look at it and still think "that's a lot of high-tech sciencey doohickamajigs right there, and this defendant is just trying to talk himself out of scientific proof! I mean, look at those knobs and needles."
It's kind of a good thing that juries are disposed to trust "sciencey" stuff, but not so good when they can't grasp what it really is they're being told about, or what the shortcomings are.
Re:Because they are unreliable. (Score:3, Insightful)
Where have I heard that before?
Re:It is better than a jury of Bobs (Score:3, Insightful)
Re:Because they are unreliable. (Score:3, Insightful)
Hey man, if you build a machine that reliably tell the difference between "right now, the subject thinks this is true" and "this is true", nobody will ever question you again - because you'll have basically constructed a universal oracle that can answer any question. You just have to find someone who believes all things are true.
Re:Lie Detection (Score:2, Insightful)
> Until they invent a machine that can travel back in time and compare the suspect's claims against the facts, there can be no lie detectors
You're talking about a "truth detector". A perfect lie detector would just need to be able to direct read access to a person's memory. While Atomicdevice has some really good points about manipulation of memory, that shouldn't discount a perfect lie detector (if one ever existed) being used in all cases. The manipulating memory thing is a problem with witnesses, not with a hypothetical perfect lie detector.
If such a detector existed, and I were accused of something I hadn't done, I'd expect to be able to use it to show that I had no recollection of commiting the act, and that I remember being somewhere else. At least I only have to worry about a jury questioning the quality of my memory rather than questioning if I'm just protecting myself. Of course, if such a thing existed, juries might start questioning why some people weren't using them...
Re:Classic statistics problem (Score:3, Insightful)