WEBVTT - Ep57 "When should new technologies enter the courtroom?" 0:00:05.040 --> 0:00:08.960 Can you measure pedophilia in a brain scan? Can you 0:00:09.000 --> 0:00:13.560 measure a lie from somebody's blood pressure? And how should 0:00:13.600 --> 0:00:16.400 a judge in court who's not an expert in science 0:00:17.040 --> 0:00:20.000 decide these things? What does any of this have to 0:00:20.000 --> 0:00:24.959 do with President Ronald Reagan or antisocial personality disorder or 0:00:25.000 --> 0:00:32.920 how the television show CSI has impacted courtrooms. Welcome to 0:00:33.000 --> 0:00:36.680 Inner Cosmos with me David Eagleman. I'm a neuroscientist and 0:00:36.720 --> 0:00:39.519 an author at Stanford and I've spent my career at 0:00:39.520 --> 0:00:43.680 the intersection of our brains and our lives. In today's 0:00:43.760 --> 0:00:49.080 episode is about an aspect of the intersection between brains 0:00:49.159 --> 0:00:57.320 and the legal system, and it's a tricky one. The 0:00:57.440 --> 0:01:04.520 question is when neuroscience techniques are allowed in courts, When 0:01:04.520 --> 0:01:08.000 should they be allowed? What bars need to be passed 0:01:08.360 --> 0:01:12.720 for a technology to be accepted. So let's start on 0:01:13.160 --> 0:01:16.640 March thirtieth, nineteen eighty one, when the President of the 0:01:16.760 --> 0:01:21.920 United States, Ronald Reagan, has been delivering a speech and afterwards, 0:01:21.920 --> 0:01:25.240 he and his team are returning to his limousine and 0:01:25.319 --> 0:01:28.600 he gives a two big arm wave to the crowd 0:01:29.160 --> 0:01:34.200 and suddenly there are gunshots ringing out and everyone's diving, 0:01:34.840 --> 0:01:38.360 and President Reagan is hit with a ricochet off his limousine, 0:01:38.680 --> 0:01:43.000 and Press Secretary James Brady falls and Secret Service agent 0:01:43.040 --> 0:01:47.280 Tim McCarthy falls, and a DC Police officer named Thomas 0:01:47.319 --> 0:01:51.520 del Haunty is also wounded, and the President arrives at 0:01:51.600 --> 0:01:56.000 the emergency room in critical condition and almost dies. And 0:01:56.080 --> 0:01:58.640 for those of you who weren't alive in nineteen eighty one, 0:01:58.960 --> 0:02:01.840 or for whom this has re seated in memory, just 0:02:01.880 --> 0:02:06.320 try to picture the horror that this entailed. Now you 0:02:06.360 --> 0:02:11.400 may remember that the gunman, John Hinckley, had a deep psychosis. 0:02:11.440 --> 0:02:15.519 He was divorced from reality, and he believed that if 0:02:15.520 --> 0:02:19.360 he shot the president, he would win the love of 0:02:19.480 --> 0:02:22.960 the actress Jody Foster. There's a lot to say about 0:02:23.000 --> 0:02:26.000 this case, and in episodes thirty six and thirty seven 0:02:26.160 --> 0:02:30.000 I talked about the insanity defense, but here I want 0:02:30.000 --> 0:02:33.400 to zoom in on a very particular aspect. The thing 0:02:33.960 --> 0:02:36.840 most salient to us today was the fact that this 0:02:37.040 --> 0:02:40.440 was the first high profile case to use a form 0:02:40.680 --> 0:02:46.280 of brain imaging. Hinckley's lawyers pled not guilty by reason 0:02:46.360 --> 0:02:51.200 of insanity, and to support their defense, they introduced brain 0:02:51.360 --> 0:02:57.799 imaging evidence so his defense counsel argued that he was schizophrenic, 0:02:57.919 --> 0:03:01.799 and they argued they could prove this by showing CAT 0:03:01.840 --> 0:03:07.120 scans or CT scans. CT stands for computer aided tomography 0:03:07.280 --> 0:03:11.760 or computerized demography. Now, the lawyers on both sides agreed 0:03:12.320 --> 0:03:16.880 that cat scans had never before been admitted as evidence 0:03:16.960 --> 0:03:20.920 in a courtroom. Neuroimaging was brand new at this time, 0:03:21.360 --> 0:03:26.440 So should the judge allow this new Fengal technology to 0:03:26.520 --> 0:03:30.880 be accepted or not? Well, it's not obvious. Can you 0:03:30.960 --> 0:03:34.760 really tell if someone suffers from schizophrenia just by looking 0:03:35.280 --> 0:03:39.640 at an anatomical picture of the brain. It's not obvious. 0:03:39.760 --> 0:03:43.240 So the judge decided to dismiss the jury so that 0:03:43.280 --> 0:03:46.360 he could hear the arguments about whether or not the 0:03:46.440 --> 0:03:51.880 technology was relevant and should be admitted. And expert witness, 0:03:51.920 --> 0:03:56.560 a physician, pointed out that Hinckley's soul sigh, which are 0:03:56.560 --> 0:03:59.880 the valleys running along the outside of the brain, these 0:04:00.080 --> 0:04:04.320 were wider than average, and this physician cited a paper 0:04:04.440 --> 0:04:10.800 suggesting a connection between schizophrenia and wider sulsie. So the 0:04:10.880 --> 0:04:15.040 assertion was, if you have schizophrenia, you can see that 0:04:15.240 --> 0:04:18.000 just by looking at a cat scan of the brain. 0:04:18.560 --> 0:04:22.400 So this doctor said, quote the fact that one third 0:04:22.440 --> 0:04:27.040 of schizophrenic participants in the study had these widened sulsie 0:04:27.320 --> 0:04:29.800 whereas in normals probably less than one out of fifty 0:04:29.839 --> 0:04:33.839 have them. That is a very powerful fact. But the 0:04:33.920 --> 0:04:37.839 prosecution rebutted this. They said, no way, it has not 0:04:38.000 --> 0:04:41.760 been proven that a cat scan can aid in the 0:04:41.839 --> 0:04:47.000 diagnosis of schizophrenia, and therefore this evidence should not be 0:04:47.080 --> 0:04:50.400 presented to the jury. In other words, they argued the 0:04:50.480 --> 0:04:54.000 technology should be excluded from the courtroom because it was 0:04:54.040 --> 0:04:57.880 not yet ready for prime time. The judge listened to 0:04:57.920 --> 0:05:01.479 the arguments and he finally decided that he would not 0:05:01.880 --> 0:05:05.640 admit the cat scan. Then nine days later he heard 0:05:05.680 --> 0:05:08.960 more expert testimony and he confirmed that he would not 0:05:09.120 --> 0:05:12.760 take the cat scan. And then he changed his mind 0:05:12.800 --> 0:05:15.800 and reported he would take the cat scan. Okay, So 0:05:15.880 --> 0:05:19.200 what is this back and forth illustrate? It illustrates the 0:05:19.360 --> 0:05:24.840 difficulty for a judge to decide what makes meaningful evidence 0:05:25.000 --> 0:05:28.880 and what does not. In the end, Hinckley was found 0:05:29.120 --> 0:05:32.240 not guilty by reason of insanity, although that had little 0:05:32.480 --> 0:05:34.760 or nothing to do with the cat scan. But this 0:05:34.880 --> 0:05:38.560 high profile case is just one of hundreds where this 0:05:38.680 --> 0:05:43.400 question comes up, should neuroimaging be allowed in the courtroom. 0:05:43.720 --> 0:05:47.080 There's no single answer to this question, and in part 0:05:47.120 --> 0:05:49.880 that's because there are many different guyses in which it 0:05:49.920 --> 0:05:52.320 comes up. And so that's what we're going to talk 0:05:52.360 --> 0:05:55.520 about today. We're going to talk about how any technology 0:05:55.920 --> 0:06:01.760 gets into courtrooms. So to motivate this, imagine that we 0:06:01.880 --> 0:06:06.280 start seeing advertisements for a new Silicon Valley company that 0:06:06.320 --> 0:06:10.599 has developed a new mind reading technology. They call this 0:06:10.800 --> 0:06:14.440 the Palo Alto three thousand, and they strap it to 0:06:14.480 --> 0:06:17.719 your head and they measure some brain waves and pass 0:06:17.839 --> 0:06:21.320 that through a large language model, and they print out 0:06:21.360 --> 0:06:24.680 in words to a screen what you are thinking. So 0:06:24.720 --> 0:06:28.200 you might be thinking about wanting a hot dog with pickles. 0:06:28.200 --> 0:06:31.440 In this machine will print to the screen, I want 0:06:31.440 --> 0:06:34.600 a hot dog with pickles. Now this is totally made up, 0:06:34.680 --> 0:06:38.120 but pretend it's true. In a few years that said 0:06:38.200 --> 0:06:42.920 company launches, and let's say the technology looks pretty good. 0:06:42.960 --> 0:06:47.040 It captures the gist of what you are thinking about. Now, 0:06:47.080 --> 0:06:51.400 the question is should this be admissible in a court 0:06:51.520 --> 0:06:55.640 of law. Let's imagine that someone puts it on and 0:06:55.720 --> 0:06:59.080 states under oath that on April twenty fifth, he was 0:06:59.080 --> 0:07:02.320 getting dinner with his family, but suddenly the screen prints 0:07:02.360 --> 0:07:05.520 out I committed the crime. Now, how do you know 0:07:05.560 --> 0:07:10.240 whether to believe that or not? The company is started 0:07:10.240 --> 0:07:12.800 by a handful of young people who dropped out of college, 0:07:13.040 --> 0:07:16.360 and they claim to be experts in neuroscience, But how 0:07:16.360 --> 0:07:20.520 do you know whether it really works? And especially in 0:07:20.560 --> 0:07:23.840 a high stakes situation, should you accept this in a 0:07:23.960 --> 0:07:27.800 court of law or not? Well, some people, in order 0:07:27.880 --> 0:07:30.600 to judge the quality of the technology, they ask, well, 0:07:30.640 --> 0:07:33.520 are they charging for this technology? But that's not a 0:07:33.760 --> 0:07:37.320 meaningful measure. Of course they're charging. They can't develop new 0:07:37.360 --> 0:07:41.560 technologies for free, anymore than you would expect Apple to 0:07:41.640 --> 0:07:44.440 not charge for their laptop. But the fact that they're 0:07:44.840 --> 0:07:49.560 charging certainly doesn't rule in or out anything about its efficacy. 0:07:49.720 --> 0:07:53.160 So how do you know whether the technology is efficacious? 0:07:53.240 --> 0:07:55.440 Can it be used in a court of law? How 0:07:55.480 --> 0:07:58.800 do you know whether it works and provides what the 0:07:58.880 --> 0:08:03.480 legal system calls probative value, which means can it do 0:08:03.560 --> 0:08:07.280 what it's supposed to do? Can it provide sufficiently useful 0:08:07.400 --> 0:08:12.080 evidence to prove something in a trial? So this is 0:08:12.120 --> 0:08:14.720 what we're going to talk about today. Most of the 0:08:14.760 --> 0:08:18.360 time we don't realize that new technologies always have to 0:08:18.400 --> 0:08:22.640 be assessed by courtrooms to know whether they should be 0:08:22.720 --> 0:08:26.320 accepted or rejected. And some get in and then we 0:08:26.360 --> 0:08:29.560 take that as background furniture, and others never make it. 0:08:29.840 --> 0:08:32.600 And what we're going to see today is how and why. 0:08:33.160 --> 0:08:37.120 So fast forward some decades from the Hinckley trial. Where 0:08:37.160 --> 0:08:40.680 are we now? What is allowed in the courtroom? Well, 0:08:40.720 --> 0:08:45.280 we have more sophisticated technologies to image the brain. Now, 0:08:45.360 --> 0:08:48.480 for example, we can get a picture of the brain 0:08:48.679 --> 0:08:53.520 in an MRII scan. Magnetic resonance imaging MRI gives you 0:08:53.679 --> 0:08:57.800 a snapshot of what the brain of a person looks like. 0:08:58.200 --> 0:09:01.679 You're not seeing activity there, You're just seeing the anatomy. 0:09:02.360 --> 0:09:04.760 Think of this an analogy to the way you would 0:09:04.760 --> 0:09:07.080 look at someone's skeleton with an X ray. You can't 0:09:07.080 --> 0:09:10.640 see anything moving around what you see as a snapshot. 0:09:10.960 --> 0:09:14.920 So with MRI you can hope to see abnormalities like 0:09:15.000 --> 0:09:19.000 a tumor or evidence of a stroke, or the consequences 0:09:19.040 --> 0:09:22.960 of a traumatic brain injury. Now, I've been called by 0:09:23.040 --> 0:09:26.160 many defense lawyers over the years who say, I have 0:09:26.200 --> 0:09:28.880 a client who's going up for trial. Can you take 0:09:28.920 --> 0:09:31.800 a brain scan and see if you can find something 0:09:31.920 --> 0:09:34.680 wrong with their brain, so this can serve as a 0:09:34.800 --> 0:09:37.680 mitigating factor. But I always tell them the same thing. 0:09:38.120 --> 0:09:40.920 If you find something wrong with your client's brain, that 0:09:40.960 --> 0:09:44.680 can serve as a double edged sword. The duray might think, Okay, 0:09:44.720 --> 0:09:48.319 I'm convinced there's something different about this man's brain. But 0:09:48.440 --> 0:09:52.920 this presumably means he'll be predisposed to committing this kind 0:09:52.960 --> 0:09:55.840 of crime again, so we should probably lock him up 0:09:55.880 --> 0:09:58.880 for a longer time. So a defense lawyer has to 0:09:59.400 --> 0:10:03.480 utilize this argument with care. In any case, what MRI 0:10:03.760 --> 0:10:07.600 gives you is an anatomical snapshot. And now I want 0:10:07.640 --> 0:10:11.280 to tell you about the next level of technology called fMRI. 0:10:11.760 --> 0:10:15.120 Where the F stands for fancy MRI. Okay, I'm kidding. 0:10:15.240 --> 0:10:20.400 It stands for functional magnetic resonance imaging fMRI. And this 0:10:20.440 --> 0:10:24.320 is because it's telling you about the function of the brain. 0:10:24.920 --> 0:10:28.920 It's measuring blood flow to show you where the activity 0:10:28.920 --> 0:10:32.480 in the brain just was. This works because when brain 0:10:32.559 --> 0:10:36.280 cells are active, they consume energy, and the blood flow 0:10:36.320 --> 0:10:39.760 to that specific region needs to increase so that you 0:10:39.800 --> 0:10:43.680 can bring fresh oxygenated blood to the area to restore 0:10:43.720 --> 0:10:48.280 the used energy. So in fMRI, we see where the 0:10:48.320 --> 0:10:51.840 new oxygenated blood is going and we say, aha, there 0:10:51.920 --> 0:10:55.360 must have just been some activity there a few seconds ago. 0:10:55.880 --> 0:10:59.320 So that's the difference between an anatomical snapshot or a 0:10:59.520 --> 0:11:03.200 functional picture of what's going on. Now. Part of the 0:11:03.240 --> 0:11:06.839 reason that you can use the static snapshot the MRI 0:11:07.160 --> 0:11:11.360 in court is because it's generally seen as hard science. 0:11:11.520 --> 0:11:15.240 This is the guy's brain. But when we're talking about fMRI, 0:11:16.000 --> 0:11:17.960 what we're looking at is the activity in the brain, 0:11:18.280 --> 0:11:22.120 and we're generally asking something about the person's mental state, 0:11:22.640 --> 0:11:26.160 and can that be the same kind of hard science. 0:11:26.559 --> 0:11:28.680 On the one hand, it's a clear question with a 0:11:28.720 --> 0:11:31.600 clear answer if someone has a stroke or a brain tumor. 0:11:32.000 --> 0:11:34.960 But this isn't the case if you want to pose 0:11:35.000 --> 0:11:39.880 a question like did this defendant intend to kill the victim? 0:11:40.320 --> 0:11:44.760 fMRI doesn't and can't give you clear answers like that 0:11:45.440 --> 0:11:48.240 to questions that are useful for the legal system. So 0:11:48.280 --> 0:12:04.080 we're going to dig into this now. So first let's 0:12:04.080 --> 0:12:08.559 start with the question of whether fMRI has been used 0:12:08.600 --> 0:12:11.920 in courts. The answer is yes, But the technology can 0:12:11.960 --> 0:12:14.160 be used in different ways. It doesn't always have to 0:12:14.200 --> 0:12:19.000 involve an individual's brain, but can sometimes be about brains 0:12:19.240 --> 0:12:22.000 in general. So let me give you an example. There 0:12:22.040 --> 0:12:25.559 was a murder case in Missouri where a young man 0:12:25.640 --> 0:12:29.240 named Christopher Simmons broke into the home of a woman 0:12:29.320 --> 0:12:33.680 named Shirley Crook. He covered her eyes and mouth with 0:12:33.800 --> 0:12:36.920 duct tape, he bound her hands together, and then he 0:12:37.040 --> 0:12:40.440 drove her to a state park and threw her off 0:12:40.559 --> 0:12:44.920 a bridge to her death. Now, this was a premeditated crime, 0:12:45.320 --> 0:12:48.600 and the evidence was overwhelming, and he admitted to the murder. 0:12:48.960 --> 0:12:52.280 So the judge and jury handed down a death sentence. 0:12:53.040 --> 0:12:57.480 But there was a complication. Christopher Simmons was only seventeen 0:12:57.559 --> 0:13:00.680 years old at the time he committed the crime. And 0:13:00.760 --> 0:13:03.200 so this case spun all the way up to the 0:13:03.280 --> 0:13:07.439 United States Supreme Court, and the question was can you 0:13:07.679 --> 0:13:11.160 execute someone who was under the age of eighteen when 0:13:11.160 --> 0:13:15.160 they committed the crime? After all, the argument goes, adolescence 0:13:15.280 --> 0:13:19.480 is characterized by poor decision making, and young people should 0:13:19.520 --> 0:13:22.960 have the chance to grow up into a different life. Well, 0:13:23.480 --> 0:13:26.120 one of the things that happened at his trial is 0:13:26.120 --> 0:13:30.880 that the Supreme Court considered fMRI evidence. Now this wasn't 0:13:30.920 --> 0:13:35.720 from Simmons's brain in particular, but from adolescence in general. 0:13:36.360 --> 0:13:40.640 The study compared young people and adults performing the same 0:13:41.120 --> 0:13:45.240 cognitive tasks, and what the researchers found, not surprisingly, is 0:13:45.280 --> 0:13:48.160 that young brains are not doing precisely the same thing 0:13:48.160 --> 0:13:53.440 as older brains. There are measurable differences. A juvenile's brain 0:13:53.760 --> 0:13:56.600 just isn't the same thing as an adult. So the 0:13:56.640 --> 0:14:01.800 Supreme Court justices saw this evidence, considered it, and presumably 0:14:01.880 --> 0:14:04.520 this is part of what led the court to conclude 0:14:04.960 --> 0:14:09.040 that it is unconstitutional to execute someone for a crime 0:14:09.080 --> 0:14:13.120 who is a minor. Now that's an example of fMRI 0:14:13.400 --> 0:14:16.000 making it into the court. It's been used in this 0:14:16.120 --> 0:14:20.320 way to compare groups of people, juveniles versus adults in 0:14:20.360 --> 0:14:24.840 this case. But things get a little trickier when you're 0:14:24.880 --> 0:14:28.680 trying to say something about an individual's brain, the brain 0:14:29.120 --> 0:14:31.520 of the one guy standing in front of the bench. 0:14:32.160 --> 0:14:34.840 So what can we and can we not say with 0:14:34.960 --> 0:14:38.200 the technology? So let's zoom in on a few examples. 0:14:38.640 --> 0:14:42.320 Many researchers and legal minds have been asking whether one 0:14:42.360 --> 0:14:47.160 can use brain imaging to diagnose whether someone has antisocial 0:14:47.280 --> 0:14:50.600 personality disorder, which is a condition in which a person 0:14:50.680 --> 0:14:55.280 has a long term pattern of manipulating, exploiting, and violating 0:14:55.320 --> 0:15:00.600 other people people with antisocial personality disorder or a SPD. 0:15:01.240 --> 0:15:06.400 They'll commit crimes, they'll flaunt rules, they'll act impulsively and aggressively, 0:15:06.960 --> 0:15:10.480 they'll lie and cheat and steal. Now, this is a 0:15:10.520 --> 0:15:15.920 condition that is massively overrepresented in the prison population. But 0:15:16.040 --> 0:15:19.560 biologically it's not obvious what it's about. There's no single 0:15:19.680 --> 0:15:23.760 gene here, and there's not a single environmental factor. It's 0:15:23.760 --> 0:15:27.840 a complicated combination. And the legal system often cares to 0:15:27.920 --> 0:15:33.600 know whether someone has ASPD or not. And so researchers 0:15:33.720 --> 0:15:36.240 started to wonder a long time ago, could you use 0:15:36.320 --> 0:15:41.880 brain imaging to determine in some clear categorical way, does 0:15:42.040 --> 0:15:46.240 this person have ASPD or not. So, in one study, 0:15:46.600 --> 0:15:51.240 researchers highlighted the brain regions that had high probability of 0:15:51.280 --> 0:15:56.600 being anatomically different between people with ASPD and those without. 0:15:56.880 --> 0:15:59.360 And you can look in the cortex what's called the 0:15:59.360 --> 0:16:02.160 gray matter, or below the core text what's called the 0:16:02.200 --> 0:16:06.320 white matter, and you can measure these small anatomical differences 0:16:06.400 --> 0:16:09.760 between those with and without. So the question arose, can 0:16:09.840 --> 0:16:13.600 you use this technology in court as a diagnostic tool 0:16:13.920 --> 0:16:18.960 to say that this person has ASPD or not? Now 0:16:19.040 --> 0:16:20.920 do you see any problems with this off the top 0:16:20.960 --> 0:16:24.200 of your head about whether this technology can be used. 0:16:24.600 --> 0:16:28.160 The problem is that all the scientific results come about 0:16:28.440 --> 0:16:32.960 from examining groups of people, like fifty people in each group, 0:16:33.320 --> 0:16:37.320 and the question is whether these group differences are strong 0:16:37.440 --> 0:16:41.840 enough to tell you about individual differences. So this is 0:16:41.880 --> 0:16:45.320 known as the group to individual problem. In other words, 0:16:45.400 --> 0:16:48.640 you have data from groups of people that can be 0:16:48.720 --> 0:16:52.920 distinguished on average, but you're trying to say something about 0:16:53.400 --> 0:16:57.280 this individual. It would be like making an accurate statement 0:16:57.360 --> 0:17:02.000 that men on average are taught than women, and then 0:17:02.360 --> 0:17:06.199 asking whether some individual, like a tall woman, could be 0:17:06.280 --> 0:17:09.639 categorized as a man because her height clocks in at 0:17:09.680 --> 0:17:12.800 the average mail. The legal system is well aware of 0:17:12.840 --> 0:17:17.800 this grouped individual problem, and so as technologies are introduced, 0:17:18.280 --> 0:17:22.040 the justice system always needs to ask how specific is 0:17:22.080 --> 0:17:24.880 this technology and how sensitive is it? Is it good 0:17:25.040 --> 0:17:30.400 enough for individual diagnosis. Brain imaging studies generally just give 0:17:30.520 --> 0:17:34.159 us group average results, and the question is whether it 0:17:34.359 --> 0:17:38.360 tells us enough or anything about the person who's standing 0:17:38.440 --> 0:17:41.720 in front of the bench right now. Now. The idea 0:17:41.960 --> 0:17:46.320 of bringing functional brain imaging to bear on questions of 0:17:46.359 --> 0:17:50.400 criminal behavior is an old one, and this grouped individual 0:17:50.440 --> 0:17:53.680 problem is just as old. For example, there was a 0:17:53.720 --> 0:17:57.520 study in nineteen ninety seven where researchers image the brain 0:17:57.720 --> 0:18:02.880 of normal participants and murderers, and they found, on average, 0:18:02.920 --> 0:18:07.119 there was less activity in the frontal lobes in murderers. 0:18:07.560 --> 0:18:09.879 So you look at the activity in the front of 0:18:09.920 --> 0:18:12.440 the brain behind the forehead and you say, hey, on average, 0:18:12.720 --> 0:18:16.280 there's less going on here in the murderer group. But 0:18:16.400 --> 0:18:19.400 you can't use this on an individual. You can't say, oh, 0:18:19.440 --> 0:18:22.160 this person has less activity, so he must have been 0:18:22.200 --> 0:18:25.280 the murderer. In other words, it has no power in 0:18:25.320 --> 0:18:28.199 a court of law. You still face the problem of 0:18:28.359 --> 0:18:32.680 trying to say anything about an individual from a group average. 0:18:33.000 --> 0:18:36.160 And so it's for reasons like this that brain imaging 0:18:36.359 --> 0:18:40.560 on individuals has not gotten very far in courtrooms. Let 0:18:40.600 --> 0:18:44.080 me give one more example. Another research group used brain 0:18:44.160 --> 0:18:49.200 imaging fMRI to see if they could identify pedophiles. They 0:18:49.240 --> 0:18:53.600 found twenty four pedophiles and thirty four controls, and they 0:18:53.680 --> 0:18:57.919 showed them images of naked men and women and boys 0:18:57.960 --> 0:19:00.480 and girls. And what they found is that they could 0:19:00.640 --> 0:19:05.359 on average separate the participants who were pedophiles from the 0:19:05.400 --> 0:19:09.399 participants who were not. In other words, the pedophilic brain 0:19:09.560 --> 0:19:13.960 shows a subtly different signature of brain activity than the 0:19:14.040 --> 0:19:17.960 non pedophilic brain when shown these pictures. It turns out 0:19:17.960 --> 0:19:22.240 that heterosexual versus homosexual seems to be distinguishable as well. 0:19:22.840 --> 0:19:25.359 So you might think that sounds quite useful for the 0:19:25.440 --> 0:19:29.840 legal system, but when scientists and legal scholars take a 0:19:29.880 --> 0:19:33.280 closer look, it's not as clear. The first question is 0:19:33.960 --> 0:19:38.520 what are these brain signals actually measuring. The assumption is 0:19:38.560 --> 0:19:42.000 that it's measuring a state of arousal, like sexual attraction, 0:19:42.560 --> 0:19:45.840 but what else might be going on? Well, the difference 0:19:45.880 --> 0:19:49.840 in brain signals could be driven by a stress response 0:19:49.960 --> 0:19:54.239 or an anxiety response by the pedophilic participants who know 0:19:54.280 --> 0:19:58.520 they're being measured. Or perhaps what you're seeing is a 0:19:58.600 --> 0:20:02.120 measure of disgust by the non pedophilic group who knows 0:20:02.200 --> 0:20:05.000 the purpose of the study and doesn't like gazing at 0:20:05.040 --> 0:20:08.280 pictures of children in this context. Or what if the 0:20:08.280 --> 0:20:12.760 pedophilic participants were just slightly more likely to avert their 0:20:12.800 --> 0:20:16.360 eyes because of shame or not wanting to get measured. 0:20:16.840 --> 0:20:20.840 That would cause a statistical difference in the brain signals 0:20:21.200 --> 0:20:24.480 and could, in theory, explain the results. So there are 0:20:24.640 --> 0:20:29.400 lots of things that could yield this brain imaging result 0:20:29.480 --> 0:20:33.080 of a difference between the two groups, beyond the hypothesis 0:20:33.080 --> 0:20:37.959 that it's just measuring arousal, So stress, anxiety, discuss shame, 0:20:38.119 --> 0:20:41.600 all these things might be what's getting measured here. And 0:20:41.680 --> 0:20:44.000 part of why this matters is because there are many 0:20:44.480 --> 0:20:48.439 brain imaging measures where it turns out it's easy to 0:20:48.560 --> 0:20:52.320 manipulate the results. So let's say you are a pedophile 0:20:52.320 --> 0:20:54.919 who doesn't want to be labeled as such. Can you 0:20:55.040 --> 0:20:58.760 purposely move your eyes whenever you see a picture of 0:20:58.800 --> 0:21:02.399 the children, and that messes up the ability of the 0:21:02.480 --> 0:21:06.080 scanner to measure something. If something can be faked or 0:21:06.119 --> 0:21:09.720 messed up, then the technology is useless. But let's say, 0:21:09.760 --> 0:21:12.240 for argument's sake, that you have a technology that can't 0:21:12.320 --> 0:21:15.240 be faked or manipulated, and that allows us to move 0:21:15.240 --> 0:21:17.600 on to the second point. Let's say you don't even 0:21:17.720 --> 0:21:20.600 care what's getting measured, like stress or anxiety or whatever. 0:21:21.000 --> 0:21:23.280 All you care to know is whether there is a 0:21:23.359 --> 0:21:27.280 neural signature that can distinguish the pedophiles from the non pedophiles, 0:21:27.480 --> 0:21:30.920 irrespective of what is causing that signal. Well, there's also 0:21:30.960 --> 0:21:33.720 a legal problem here, which is that it's not illegal 0:21:33.800 --> 0:21:36.760 for a person to be attracted to children. It is 0:21:36.880 --> 0:21:40.840 only illegal if they act on that. All that's illegal 0:21:40.960 --> 0:21:43.840 is whether you have committed a crime or not, not 0:21:44.080 --> 0:21:47.360 whether you are attracted to children. So you can think 0:21:47.359 --> 0:21:51.720 about whatever attracts you ostriches or jello or whatever, as 0:21:51.840 --> 0:21:57.160 long as you don't commit an illegal act. So whether 0:21:57.200 --> 0:22:00.760 you're talking about ASPD or murderers or pedophile while, you'll 0:22:00.800 --> 0:22:04.159 see that measuring something that matters for a court of 0:22:04.240 --> 0:22:08.959 law isn't as straightforward as it might have originally seemed. 0:22:09.760 --> 0:22:12.800 So now let's return to the Palo Alto three thousand. 0:22:13.480 --> 0:22:18.080 The question is just because the company claims that it functions, well, 0:22:18.560 --> 0:22:20.679 how do you know whether or not to admit it 0:22:20.720 --> 0:22:23.720 into the courtroom? After all, remember what I said about 0:22:23.760 --> 0:22:27.320 the John Hinckley case, how cat scans were admitted into 0:22:27.320 --> 0:22:31.000 the court to argue that he had schizophrenia. Well, it's 0:22:31.080 --> 0:22:34.480 now known that wide and salsa in the brain have 0:22:34.600 --> 0:22:39.639 no relationship to schizophrenia. There are other better anatomical signatures 0:22:39.640 --> 0:22:42.560 that we have now, like thinner cortices in the frontal 0:22:42.600 --> 0:22:46.960 and temporal lobes and shrunkened thalamuses. But it turned out 0:22:47.400 --> 0:22:49.880 that the idea of white and salsa I just didn't 0:22:49.920 --> 0:22:52.720 hold up. Now, there was nothing fraudulent going on with 0:22:52.800 --> 0:22:55.600 the claim. It was just a new technology at the 0:22:55.680 --> 0:22:58.000 time and they were doing the best they could with 0:22:58.080 --> 0:23:01.960 small sample sizes. But it turned out the theory of 0:23:02.080 --> 0:23:05.840 white and sul sight was scientifically unsound. Remember how I 0:23:06.000 --> 0:23:09.040 mentioned that the judge went back and forth several times 0:23:09.440 --> 0:23:12.720 about the issue of whether to accept Hinckley's cat scan 0:23:12.840 --> 0:23:16.000 into the courtroom. That's exactly the right thing that should 0:23:16.040 --> 0:23:19.240 have happened. Not all claims are going to be correct 0:23:19.320 --> 0:23:24.000 just because a scientist says so. Despite best efforts, science 0:23:24.080 --> 0:23:28.359 can often be incorrect, and that is the importance of 0:23:28.440 --> 0:23:32.360 the scientific method. It's always knocking down its own walls. 0:23:33.440 --> 0:23:38.320 So what is a court to do about all this? Well, 0:23:38.800 --> 0:23:41.639 let's say that someone wants to introduce the Palo Alto 0:23:41.720 --> 0:23:46.400 three thousand into a court case, and you are the judge. 0:23:47.040 --> 0:23:49.240 You have expertise in the legal system, but you don't 0:23:49.240 --> 0:23:53.840 know the details of what's possible in neuroscience and large 0:23:53.920 --> 0:23:58.000 language models, and you have questions about whether this technology 0:23:58.080 --> 0:24:02.040 should be admitted, questions about whether it can accurately read 0:24:02.119 --> 0:24:05.760 people's thoughts, So how do you decide whether it should 0:24:05.920 --> 0:24:22.040 or should not be admitted. So let's step back to 0:24:22.119 --> 0:24:25.560 nineteen twenty three. There was a man named mister Fry 0:24:25.760 --> 0:24:29.320 who said that he had developed a lie detection technology 0:24:29.800 --> 0:24:32.800 and it relied on a measure of your blood pressure, 0:24:32.920 --> 0:24:36.480 and he wanted to introduce this into a court case 0:24:36.960 --> 0:24:38.960 the way you might want to get the Palo Alto 0:24:38.960 --> 0:24:41.600 three thousand into a case. But it turned out that 0:24:41.640 --> 0:24:46.240 mister Fry's claims were not widely accepted by anyone else 0:24:46.320 --> 0:24:50.800 in the scientific community, and so on those grounds, the 0:24:50.920 --> 0:24:55.000 court decided not to admit it into the courtroom. What 0:24:55.080 --> 0:24:59.720 they said was, look, we'll accept expert testimony that comes 0:24:59.760 --> 0:25:04.119 from well recognized science, but if there's some new technology, 0:25:04.200 --> 0:25:07.680 it has to be sufficiently established so that it's gained 0:25:08.400 --> 0:25:12.560 general acceptance in the field in which it belongs. In 0:25:12.600 --> 0:25:16.720 other words, if other experts in the field don't believe 0:25:16.800 --> 0:25:20.520 that mister Fry's systolic blood measurement is actually good at 0:25:20.520 --> 0:25:24.720 detecting lies, then you can't admit it as evidence in 0:25:24.720 --> 0:25:27.560 the court. And that case set the bar for what 0:25:27.720 --> 0:25:31.320 came to be known as the Fry standard, which is 0:25:31.320 --> 0:25:35.360 that technologies need to be generally accepted by other experts 0:25:35.400 --> 0:25:38.879 in the field before they can be admitted into the courtroom. 0:25:39.119 --> 0:25:41.919 So under the Fry standard, the court would work to 0:25:42.000 --> 0:25:46.320 determine whether the Palo Alto three thousand has met the 0:25:46.520 --> 0:25:51.520 general acceptance of the scientific community. If science experts around 0:25:51.520 --> 0:25:54.359 the world say, I've never heard of this Palo Alto 0:25:54.400 --> 0:25:56.960 three thousand, I don't think that it can actually work, 0:25:57.359 --> 0:26:01.639 then you, as the judge, can glued it from admissibility. 0:26:02.040 --> 0:26:05.800 So the court solves the problem by deferring to the 0:26:05.840 --> 0:26:09.680 expertise of other people in the field. But this isn't 0:26:09.720 --> 0:26:12.920 the only way to make that decision. The Fry standard 0:26:13.160 --> 0:26:16.320 still is the rule in about half the states in America, 0:26:16.600 --> 0:26:20.439 but the rest use a different rule to decide whether 0:26:20.600 --> 0:26:25.119 evidence should be admitted, and this is called the Dowbert standard. 0:26:25.520 --> 0:26:27.880 So in nineteen ninety three there was a lawsuit from 0:26:27.880 --> 0:26:31.960 a man named Jason Dalbert. He was born with severe 0:26:32.359 --> 0:26:36.600 birth defects and his parents brought suit against Merrill Dow 0:26:36.760 --> 0:26:41.200