WEBVTT - Ep49 "Can you read the brain to detect a lie? 0:00:05.080 --> 0:00:08.399 Can you get convicted in a court of law based 0:00:08.480 --> 0:00:13.920 on your brain activity? Our brain scan lie detectors going 0:00:14.000 --> 0:00:18.640 to be accepted? And when does measuring somebody's brain count 0:00:18.800 --> 0:00:22.800 as illegal search and seizure? When does it violate the 0:00:22.960 --> 0:00:28.080 privacy of the mind? What constitutional issues are triggered by 0:00:28.080 --> 0:00:30.840 all this? And what does this have to do with 0:00:31.080 --> 0:00:34.960 whether your mouth gets dry when you lie, or the 0:00:35.120 --> 0:00:39.040 orbits under your eyes get hot, or your voice constricts, 0:00:39.600 --> 0:00:42.040 and what the networks in your brain are up to 0:00:42.760 --> 0:00:50.760 when you deceive. Welcome to Inner Cosmos with me David Eagleman. 0:00:51.080 --> 0:00:54.200 I'm a neuroscientist and author at Stanford and in these 0:00:54.240 --> 0:00:58.200 episodes we sail deeply into our three pound universe to 0:00:58.320 --> 0:01:02.120 understand why and how our lives look the way they do. 0:01:11.400 --> 0:01:16.319 Today's episode is about the fascinating topic of whether we 0:01:16.440 --> 0:01:22.080 can use technology to figure out whether a person is lying. 0:01:22.760 --> 0:01:26.520 What is lie detection actually measuring? How good is it? 0:01:26.640 --> 0:01:29.679 When is it accepted in courts of law? And as 0:01:29.680 --> 0:01:33.760 we come to have better and better brain reading technology, 0:01:34.319 --> 0:01:38.400 what is the future of lie detection? So in a 0:01:38.440 --> 0:01:41.240 court of law, someone says I didn't do it. I 0:01:41.280 --> 0:01:44.240 wasn't even there that night, I was in my apartment 0:01:44.319 --> 0:01:48.800 watching TV. If there's no other information that can be gathered, 0:01:49.360 --> 0:01:51.600 like his cell phone is off, so there's no GPS 0:01:51.680 --> 0:01:55.960 signal and there's no eyewitnesses, there's no alimis. If there's 0:01:56.040 --> 0:02:00.760 no other information besides his testimony, how do you know 0:02:00.840 --> 0:02:03.880 if that's really what happened or not? Is he telling 0:02:03.960 --> 0:02:08.720 the truth or is he lying? Now? On every crime 0:02:08.800 --> 0:02:12.760 television show, there are lots of clever clues that are 0:02:12.880 --> 0:02:17.400 surfaced by the industrious detective. But the issue in real 0:02:17.480 --> 0:02:20.200 courts of law most of the time is that other 0:02:20.360 --> 0:02:24.960 data just isn't available, so you only have the testimony 0:02:25.080 --> 0:02:29.040 of the people who were involved. Most crime doesn't happen 0:02:29.480 --> 0:02:33.639 in fancy office buildings with full video coverage, but instead, 0:02:33.840 --> 0:02:37.800 more commonly happens in areas without any coverage, and so 0:02:38.320 --> 0:02:42.079 you need to ask people probing questions about whether they 0:02:42.120 --> 0:02:46.600 were involved and what happened, and given human nature, you 0:02:46.720 --> 0:02:49.520 really want to know whether they are telling the truth 0:02:49.800 --> 0:02:55.160 or not. So let's start today's episode with two legal cases. 0:02:55.760 --> 0:02:59.840 One involves a woman named Aditi Sharma. She moved to 0:03:00.080 --> 0:03:03.000 a town in India with her boyfriend Udit. This was 0:03:03.080 --> 0:03:06.600 an arranged marriage they were going to be together forever. 0:03:07.280 --> 0:03:09.600 But while she was there, she met a new man 0:03:09.880 --> 0:03:12.720 named Provene, and they fell in love with each other 0:03:12.880 --> 0:03:16.079 and decided that they wanted to be together rather than 0:03:16.240 --> 0:03:20.280 her and Udit being together. So she ran away with Provene, 0:03:20.800 --> 0:03:25.399 but Udit kept following them, so she and Pravene decided 0:03:25.440 --> 0:03:28.560 that they would murder Udit, and they did so by 0:03:28.600 --> 0:03:31.520 inviting him to meet them Adam McDonald's, and they gave 0:03:31.600 --> 0:03:35.080 him a popular drink in India that was laced with 0:03:35.360 --> 0:03:38.520 arsenic and he drank it and died. So she was 0:03:38.520 --> 0:03:42.120 brought up on charges of murder, and in court she 0:03:42.320 --> 0:03:45.960 denied the crime. But part of what led her murder 0:03:46.040 --> 0:03:50.360 conviction was a lie detection test what's called an electro 0:03:50.480 --> 0:03:54.840 encephalogram or EEG. So we'll come back to this in 0:03:54.880 --> 0:03:57.480 a few moments now. Case number two is in the 0:03:57.560 --> 0:04:01.920 United States, a case in Tennessee. A man named Semrau 0:04:02.200 --> 0:04:05.640 was the CEO of two nursing home facilities, and these 0:04:05.640 --> 0:04:11.000 facilities were accused of having the employees fraudulently fill out 0:04:11.080 --> 0:04:15.120 Medicare and Medicaid forms, and Semrau said, I had no 0:04:15.280 --> 0:04:17.599 knowledge of this. I did not know this was happening 0:04:17.640 --> 0:04:21.279 in my facilities. And so in going to court, he 0:04:21.480 --> 0:04:25.279 enlisted the help of a lie detection company that used 0:04:25.680 --> 0:04:30.120 new brain scan lie detection. This is a new technology 0:04:30.120 --> 0:04:32.800 that we'll talk about today. So this company did the 0:04:32.800 --> 0:04:37.000 brain scan and the results suggested that in fact Semrau 0:04:37.240 --> 0:04:40.479 was not lying, that he had no knowledge of the fraud. 0:04:40.720 --> 0:04:43.479 And so he went to court and the question is 0:04:43.839 --> 0:04:47.719 did the court accept the brain scan? The answer is 0:04:47.839 --> 0:04:50.279 they did not. They wouldn't accept in court. They said, 0:04:50.640 --> 0:04:54.120 you can't present that here. So should they have accepted it? 0:04:54.440 --> 0:04:57.279 How do we decide when to accept a new technology 0:04:57.360 --> 0:05:01.440 or not? Under what circumstances do you say this technology 0:05:01.560 --> 0:05:03.880 is good enough that it's telling us what we need 0:05:03.920 --> 0:05:06.080 to know. So these are all the issues that we're 0:05:06.080 --> 0:05:08.479 going to talk about today, and we'll get into the 0:05:08.520 --> 0:05:11.480 background of what a lie is and how you could 0:05:11.520 --> 0:05:14.400 measure something about it and where this is all going. 0:05:15.279 --> 0:05:18.000 So we'll start with the basic question of what is 0:05:18.120 --> 0:05:22.000 a lie. Well, first, lying might seem like a human invention. 0:05:22.200 --> 0:05:26.960 So it's interesting to know that we see deception throughout 0:05:27.000 --> 0:05:30.720 the kingdom of life. For example, in the world of plants, 0:05:31.160 --> 0:05:34.240 there's a type of orchid that develops to look just 0:05:34.360 --> 0:05:38.240 like a female bee, and this way male bees come 0:05:38.279 --> 0:05:41.120 and land on it, and then the plant eats them. 0:05:41.440 --> 0:05:45.000 It's a strategy of deception to lure in the bee, 0:05:45.480 --> 0:05:48.320 and we see this throughout the animal kingdom. For example, 0:05:48.839 --> 0:05:53.200 fireflies blink to say, hey, I'm here, please meet with me. 0:05:53.880 --> 0:05:58.520 And there's a predatory firefly beetle named Fotaurus, and this 0:05:58.680 --> 0:06:04.880 meetle has evolved bioluminescence so that it flashes light signals 0:06:04.880 --> 0:06:09.279 of other firefly species, and that attracts the males and 0:06:09.320 --> 0:06:12.200 then they kill and eat the males. So this is 0:06:12.240 --> 0:06:16.880 known as aggressive mimicry. They're pretending they're the other fireflies. 0:06:17.000 --> 0:06:18.840 So one way we can think about this is that 0:06:18.920 --> 0:06:23.720 they are lying. They're not displaying truthy signals about who 0:06:23.720 --> 0:06:27.559 they are, but instead they are deceiving to accomplish their ends. 0:06:28.000 --> 0:06:31.480 So deception is not simply a human thing. But let's 0:06:31.520 --> 0:06:34.320 turn to humans because that's what we care about. So 0:06:34.480 --> 0:06:38.680 the first question is how do we define a lie? Well, 0:06:39.080 --> 0:06:41.880 it turns out that's not so easy. Let's consider an 0:06:41.920 --> 0:06:46.760 example like the scientist Copernicus. He developed a mathematical model 0:06:46.800 --> 0:06:49.240 in which the sun goes around the earth, and that 0:06:49.320 --> 0:06:52.400 turns out to be totally wrong. But he wasn't lying, 0:06:52.520 --> 0:06:57.000 he wasn't intending to deceive. And occasionally what you'll find 0:06:57.080 --> 0:07:00.599 is a situation where someone tells a man this truth, 0:07:00.920 --> 0:07:03.000 but they had no intention to deceive. I'll give you 0:07:03.080 --> 0:07:05.880 an example. Some years ago, there was a Harvard kid 0:07:05.880 --> 0:07:08.800 who went out and drank too much, and as he 0:07:08.880 --> 0:07:12.440 was stumbling home, he was jumped by two locals who 0:07:12.480 --> 0:07:15.960 beat him up badly and smashed his head into the sidewalk. 0:07:16.120 --> 0:07:19.720 So in an effort to save himself, he pulled a 0:07:19.760 --> 0:07:22.840 pen knife out of his pocket and he stabbed at 0:07:22.840 --> 0:07:25.840 the assailants to get them away. So they ran away 0:07:26.320 --> 0:07:29.600 seemingly unharmed. But it turns out his knife had gone 0:07:29.960 --> 0:07:33.360 through the chest of one assailant and tore the lining 0:07:33.400 --> 0:07:37.160 around the heart, causing what's called a cardiac tamponaud in 0:07:37.200 --> 0:07:40.560 which blood fills the lining around the heart and squishes it, 0:07:40.840 --> 0:07:45.920 and within hours this assailant died. Now, when the police 0:07:46.040 --> 0:07:49.520 came to question this Harvard kid, he told them a 0:07:49.640 --> 0:07:54.160 totally untrue story about what had happened. Why, While a 0:07:54.240 --> 0:07:58.000 neurologist and researcher named Jeremy Schmahman was called in to 0:07:58.160 --> 0:08:01.360 diagnose and testify, and he realized it was because this 0:08:01.480 --> 0:08:05.000 young man had been badly concussed, His head had been 0:08:05.040 --> 0:08:09.160 pounded into the sidewalk, and his brain had constructed a 0:08:09.280 --> 0:08:13.520 false memory about the whole event. So the opposing lawyer asked, 0:08:13.920 --> 0:08:18.360 do concussions make you lie? But Schmaman explained that a 0:08:18.520 --> 0:08:23.840 patient with a concussion is not lying, he's confabulating. There's 0:08:23.840 --> 0:08:28.800 a critical difference there because the patient believes it. In 0:08:28.840 --> 0:08:32.959 other words, just because a statement is not factual doesn't 0:08:33.000 --> 0:08:35.880 make it a lie. And in a future episode, I'm 0:08:35.920 --> 0:08:40.000 going to dive really deeply into this issue about confabulation 0:08:40.240 --> 0:08:43.560 because we see this all the time in many different 0:08:43.600 --> 0:08:47.040 types of brain damage. This even happened to a Supreme 0:08:47.120 --> 0:08:50.959 Court Justice, William Douglas, who ended up with a disorder 0:08:51.040 --> 0:08:54.800 called a nosygnosia in which he lost the ability to 0:08:54.920 --> 0:08:58.319 control part of his body, and yet he spoke falsely 0:08:58.400 --> 0:09:02.280 about it. The key is he was lying, he was confabulating. 0:09:02.360 --> 0:09:05.880 He believed that he could still move his body. So 0:09:06.240 --> 0:09:09.720 there are many situations where a person can believe something 0:09:09.920 --> 0:09:14.480 falsely and that is not equivalent to a lie. And 0:09:14.559 --> 0:09:16.560 of course, this comes up all the time when people 0:09:16.600 --> 0:09:21.080 are asked about their memory of an event, because memory 0:09:21.520 --> 0:09:25.440 doesn't store sequences of zeros and ones like a computer does. 0:09:25.920 --> 0:09:28.679 And I talked about this at length in the context 0:09:28.679 --> 0:09:32.360 of eyewitness testimony in episode twenty seven. We are not 0:09:32.760 --> 0:09:36.040 like a video recorder, and my memory of something that 0:09:36.080 --> 0:09:40.320 went down might genuinely be very different from your memory. 0:09:40.480 --> 0:09:42.760 But again, if we both probe our memories in the 0:09:42.800 --> 0:09:46.160 court of law, and we have different versions, it doesn't 0:09:46.200 --> 0:09:51.199 require that one of us is purposefully lying. But let's 0:09:51.240 --> 0:09:57.160 turn to actual deceptive lies, where you're intending to fool somebody. 0:09:57.200 --> 0:10:02.080 What do we know about that? Welltive lying starts in 0:10:02.480 --> 0:10:06.480 children at about the age of five, they start to lie. 0:10:06.600 --> 0:10:11.520 It's a cognitive development. When you're very young, you believe 0:10:11.559 --> 0:10:14.760 that your parents know everything. But around five years into 0:10:14.800 --> 0:10:18.079 the game, you realize that your parents are not omniscient, 0:10:18.160 --> 0:10:20.960 they don't know everything that happened, and it strikes you 0:10:21.040 --> 0:10:24.920 that you can inject false information. And so we train 0:10:25.000 --> 0:10:27.800 our youth on the value of truth telling because this 0:10:27.880 --> 0:10:31.680 is very pro social behavior. But the fact is, when 0:10:31.679 --> 0:10:36.080 the stakes are high, humans will often continue to practice deception, 0:10:36.240 --> 0:10:40.840 perhaps in small ways throughout their lives. Now, interestingly, we 0:10:40.920 --> 0:10:44.120 see that other primates do the same thing. They do 0:10:44.200 --> 0:10:48.080 what's called tactical deception. And the general story we find 0:10:48.120 --> 0:10:52.600 across primate species is that the bigger the brain they have, 0:10:53.240 --> 0:10:57.240 the more deception they do, because bigger brains allow us 0:10:57.320 --> 0:11:01.920 to model what other brains are thinking, which allows us 0:11:01.960 --> 0:11:05.199 to be better at manipulating them. But what's interesting is 0:11:05.240 --> 0:11:09.440 that understanding what other brains are thinking is also the 0:11:09.520 --> 0:11:14.320 development that allows species to build larger social groups with 0:11:14.600 --> 0:11:19.160 richer collaborations. So we find this strange truth that the 0:11:19.200 --> 0:11:24.040 same evolution that allows us to live in groups harmoniously 0:11:24.600 --> 0:11:28.920 also allows us to deceive. Okay, so you might think, well, 0:11:29.000 --> 0:11:31.160 we have these big brains and we're able to lie. 0:11:31.320 --> 0:11:34.040 But you might think, how could you possibly build a 0:11:34.200 --> 0:11:38.640 technology that can tell whether something is a truth or untruth. 0:11:39.080 --> 0:11:44.160 It all comes down to a simple principle. Lying is hard. 0:11:44.600 --> 0:11:48.199 Why because you have to think of the true testimony 0:11:48.520 --> 0:11:51.200 and inhibit that so you don't blurt it out, and 0:11:51.240 --> 0:11:54.920 then you have to generate a plausible alternative, and all 0:11:54.960 --> 0:11:56.720 the while you have to do a good job of 0:11:56.880 --> 0:12:01.760 hiding your anxiety and guilt. You have to manage your emotions. 0:12:02.000 --> 0:12:05.120 You have to inhibit any behaviors that would give you away. 0:12:05.200 --> 0:12:09.559 You might try to generate honest looking behaviors. So that's 0:12:09.559 --> 0:12:11.280 what we're going to talk about today. How do you 0:12:11.360 --> 0:12:14.920 build a technology to detect all of that extra work 0:12:15.200 --> 0:12:33.560 that goes into a lie. Now, what you might find 0:12:33.640 --> 0:12:37.880 unexpected is that the idea of lie detection of using 0:12:37.880 --> 0:12:41.360 a technology in the courtroom, this has a long history. 0:12:41.559 --> 0:12:44.840 So take a guess when the first lie detector was 0:12:44.920 --> 0:12:50.080 used in courts two thousand years ago in China, and 0:12:50.120 --> 0:12:56.240 the technology was uncooked rice. The defendant would put uncooked 0:12:56.280 --> 0:12:59.920 rice in his mouth and he was then asked questions. Now, 0:13:00.320 --> 0:13:04.440 presumably his answers were marginally more difficult to understand, but 0:13:04.480 --> 0:13:07.959 that wasn't the important part. The critical test was whether 0:13:08.000 --> 0:13:10.680 he could spit the rice back out of his mouth. 0:13:11.080 --> 0:13:14.760 If he couldn't, it was assumed that his dry mouth 0:13:15.320 --> 0:13:19.640 served as a physiologic marker of his nervousness about lying, 0:13:20.040 --> 0:13:21.959 And the idea was that if they were telling the truth, 0:13:22.200 --> 0:13:24.360 they'd have plenty of saliva in their mouth and they'd 0:13:24.360 --> 0:13:27.280 be able to spit the rice out. So this was 0:13:27.320 --> 0:13:31.480 the first lie detector test. But Eventually, starting about a 0:13:31.520 --> 0:13:35.360 century ago, humans leveled this up. They invented what's commonly 0:13:35.400 --> 0:13:39.439 known as the polygraph test, which is usually what people 0:13:39.480 --> 0:13:42.760 refer to when they talk about a lie detector. Now, 0:13:42.760 --> 0:13:45.520 what you do is you attach some sensors to a 0:13:45.559 --> 0:13:50.439 person's hand, and you're measuring their physiologic responses while they're 0:13:50.480 --> 0:13:54.000 answering questions. You're measuring their heart rate, their blood pressure, 0:13:54.240 --> 0:13:58.680 their breathing, their skin conductance. And the underlying principle is 0:13:58.679 --> 0:14:03.600 that when a person lives, their body reveals changes because 0:14:03.600 --> 0:14:08.200 of the stress and anxiety associated with deception, and so 0:14:08.320 --> 0:14:12.960 the examiner analyzes the recordings to look for patterns or 0:14:13.000 --> 0:14:17.480 deviations that indicate lying. So to make this clear, let's 0:14:17.559 --> 0:14:21.080 zoom in on the issue of skin conductance. So you've 0:14:21.120 --> 0:14:24.000 got all these little sweat glands in your skin, and 0:14:24.040 --> 0:14:27.760 when you have a sudden stress response, those sweat glands 0:14:27.800 --> 0:14:30.600 open more, which would cause you to sweat more. But 0:14:30.640 --> 0:14:33.440 it's not sweat that's being measured here. What's being measured 0:14:34.200 --> 0:14:37.720 is that there's a slight electrical current between point A 0:14:37.880 --> 0:14:42.120 and point B across your skin, and the conductance between 0:14:42.200 --> 0:14:46.800 A and B changes when those sweat glands open, so 0:14:46.960 --> 0:14:51.560 the electrical conductance across your skin, or inversely the resistance 0:14:52.120 --> 0:14:55.520 that changes. And so the measure here is just looking 0:14:55.560 --> 0:14:59.240 at how well electricity can pass from here to there 0:14:59.280 --> 0:15:03.000 across your skin. And when you are stressed, the electricity 0:15:03.040 --> 0:15:05.800 passes better. So you get a blip on the graph 0:15:06.160 --> 0:15:09.840 when you have a sudden moment of stress. Now here's 0:15:09.880 --> 0:15:13.960 the critical point. It's not measuring a lie, it's just 0:15:14.000 --> 0:15:18.200 measuring the stress response that's associated with lying. You're looking 0:15:18.240 --> 0:15:22.800 for that stress response as a correlate of deception. So 0:15:23.040 --> 0:15:25.640 you ask me, were you at this location on the 0:15:25.720 --> 0:15:28.880 night of March first, and I say no, I wasn't. 0:15:29.160 --> 0:15:32.400 Now if I actually was there, the lie is stressful, 0:15:32.480 --> 0:15:35.200 even at a subconscious level, and that can be read 0:15:35.240 --> 0:15:38.840 out because I have an automatic under the hood stress 0:15:38.920 --> 0:15:43.760 response that registers. The sweat glands on my skin open slightly, 0:15:44.000 --> 0:15:48.560 which changes the electrical conductance across my skin, Maybe my 0:15:48.680 --> 0:15:51.480 breathing changes a bit, my heart rate changes slightly, and 0:15:51.520 --> 0:15:55.880 as a result, this deception is brought to light. Now, 0:15:56.560 --> 0:15:59.280 this is the common way to use the polygraph test, 0:15:59.320 --> 0:16:01.920 but there's also second way that it's used, and that's 0:16:02.040 --> 0:16:06.200 known as the guilty knowledge technique. It turns out that 0:16:06.280 --> 0:16:10.200 if I show you something that's familiar to you versus unfamiliar, 0:16:10.680 --> 0:16:15.760 your brain will have a different unconscious response. So imagine 0:16:15.800 --> 0:16:18.760 I show you something like a strange little three D 0:16:19.000 --> 0:16:22.160 statue of a Pokemon. So if it's the first time 0:16:22.200 --> 0:16:25.560 you've ever seen that statue, your brain has a different 0:16:25.600 --> 0:16:28.920 response to it than if it's something you've seen before. 0:16:29.720 --> 0:16:33.560 So the police or the private investigator can leverage the 0:16:33.680 --> 0:16:37.360 guilty knowledge technique because let's say they know that this 0:16:38.000 --> 0:16:41.280 unique little statue of the Pokemon was at the scene 0:16:41.280 --> 0:16:43.720 of the crime. It was on the desk of the 0:16:43.760 --> 0:16:46.520 man who got murdered. So they sit you down in 0:16:46.560 --> 0:16:49.720 the chair and they put the polygraph electrodes on your skin, 0:16:50.240 --> 0:16:52.960 and then they show you a can of coke, and 0:16:52.960 --> 0:16:54.920 then they show you a computer printer, and then they 0:16:54.920 --> 0:16:57.480 show you a large spoon, and then they show you 0:16:57.880 --> 0:17:02.240 the Pokemon statue. Woh big response. Your brain has a 0:17:02.280 --> 0:17:05.880 different response because you were there, and only somebody who 0:17:05.880 --> 0:17:09.520 had seen that statue before and therefore would have been 0:17:09.520 --> 0:17:13.600 at the scene of the crime, would have that physiologic response. 0:17:14.040 --> 0:17:19.560 You're looking for a familiarity response for something that only 0:17:19.600 --> 0:17:22.680 a person who was there would know. That's the guilty 0:17:22.800 --> 0:17:26.960 knowledge technique. It's also known as the concealed information test. 0:17:27.560 --> 0:17:29.800 And keep in mind that your brain picks up all 0:17:29.920 --> 0:17:31.960 kinds of things that you're not even aware of. So 0:17:32.000 --> 0:17:36.560 it turns out that these familiarity responses can be big, 0:17:36.680 --> 0:17:39.679 whether or not you know that you have seen that 0:17:39.720 --> 0:17:43.240 thing before. So the polygraph test is used in both 0:17:43.280 --> 0:17:48.080 these ways, judging stress or judging familiarity. So let's go 0:17:48.160 --> 0:17:51.840 back to nineteen thirty two when a couple got married 0:17:52.000 --> 0:17:56.920 in the Northwestern Crime Scene Laboratory under the polygraph test. 0:17:57.160 --> 0:17:59.920 This couple got married and the newspapers at the time 0:18:00.119 --> 0:18:03.560 reported that her heart almost stopped beating when she said 0:18:03.600 --> 0:18:06.440 I do, and he twitched a little when he said 0:18:06.480 --> 0:18:08.800 I do. And the papers reported that they thought this 0:18:08.880 --> 0:18:12.000 all looked fine and good, but of course people didn't 0:18:12.160 --> 0:18:14.760 typically get married while hooked up to this test. This 0:18:14.960 --> 0:18:17.879 was a publicity stunt because the groom was one of 0:18:17.880 --> 0:18:21.040 the founders of the polygraph test, and he had launched 0:18:21.240 --> 0:18:25.000 a polygraph testing company and he was jazzed to get 0:18:25.040 --> 0:18:30.040 this pr But let's dive into its usage. So it 0:18:30.119 --> 0:18:33.440 was first used in a criminal case in nineteen twenty 0:18:33.480 --> 0:18:37.960 three in Berkeley, California, after a retired police chief was murdered, 0:18:38.440 --> 0:18:41.640 and then it started gaining traction in law enforcement all 0:18:41.680 --> 0:18:44.040 through the twenties and thirties, and then by the time 0:18:44.080 --> 0:18:46.600 of the Cold War, there was a lot of concern 0:18:46.680 --> 0:18:50.840 about possible espionage, so a lot of agencies started to 0:18:50.960 --> 0:18:54.800 use the polygraph test for security screenings, and by the 0:18:54.880 --> 0:18:59.560 nineteen sixties, essentially all federal law enforcement agencies were leveraging 0:18:59.600 --> 0:19:03.640 this test. But then by the nineteen seventies some academics 0:19:03.680 --> 0:19:07.800 started raising questions about the accuracy of the polygraph test, 0:19:07.840 --> 0:19:11.000 because it's quite good, but it's not perfect, and so 0:19:11.280 --> 0:19:15.000 battles over its admissibility, and courts have raged since at 0:19:15.080 --> 0:19:17.520 least that time. So I'll give you a sense of 0:19:17.560 --> 0:19:19.960 how these battles can go. In nineteen ninety two, there 0:19:20.040 --> 0:19:23.360 was a US Air Force man named Edward Scheffer who 0:19:23.560 --> 0:19:27.000 applied to work at the Office of Special Investigations, and 0:19:27.080 --> 0:19:30.879 they said, cool, but you have to take a urinalysis 0:19:31.080 --> 0:19:34.080 and a polygraph test to make sure you're not on drugs. 0:19:34.400 --> 0:19:36.960 So he said, great, and he did that. What happened 0:19:37.040 --> 0:19:42.720 is that the uranalysis test came back positive for methamphetamine use. 0:19:43.240 --> 0:19:46.840 But his polygraph test in which he said I have 0:19:46.960 --> 0:19:49.480 not done any drugs since I joined the Air Force, 0:19:49.760 --> 0:19:53.199 that came back and showed no evidence of deception, in 0:19:53.240 --> 0:19:57.080 other words, showing he was not lying. Now, based just 0:19:57.160 --> 0:20:01.560 on the uranalysis, he was tried by Marshal for using 0:20:01.640 --> 0:20:04.520 drugs and at the court martial he said, yeah, I know, 0:20:04.600 --> 0:20:07.639 my year analysis came back positive, but I would like 0:20:07.720 --> 0:20:10.720 to submit my polygraph test as part of my defense 0:20:10.800 --> 0:20:14.320 because that indicates I am not lying. And the court 0:20:14.359 --> 0:20:17.920 martial judge said sorry, but as a rule, the Air 0:20:17.960 --> 0:20:22.520 Force never accepts polygraph tests in the courtroom. And Scheffer said, 0:20:23.080 --> 0:20:26.720 this violates my sixth Amendment rights, which say that I 0:20:26.760 --> 0:20:29.879 should be able to submit evidence in my defense, and 0:20:29.880 --> 0:20:32.880 you're not letting me do this. But the military court 0:20:32.880 --> 0:20:36.679 wouldn't accept the polygraph and so Scheffer was convicted and 0:20:36.720 --> 0:20:39.680 dishonorably discharged. And so he went to the Air Force 0:20:39.720 --> 0:20:42.160 Court of Appeals and they said, yeah, no, we don't 0:20:42.160 --> 0:20:44.800 have to take the polygraph test, even if it would 0:20:44.880 --> 0:20:47.880 exculpate you. So then he went to the Armed Forces 0:20:47.920 --> 0:20:50.920 Court of Appeals and they said, you know you're right, 0:20:51.000 --> 0:20:55.199 this does seem to violate your sixth Amendment right to 0:20:55.400 --> 0:20:58.359 present a defense. So this went all the way up 0:20:58.359 --> 0:21:01.679 to the Supreme Court, and they had to struggle with 0:21:01.760 --> 0:21:05.840 this because the polygraph test is controversial in many ways, 0:21:06.400 --> 0:21:08.560 and so what they decided is that they were going 0:21:08.640 --> 0:21:11.639 to side with the military on this. The military courts, 0:21:11.680 --> 0:21:15.200 they said, do not have to take polygraph evidence if 0:21:15.240 --> 0:21:18.560 they don't want to. They said, we need to preserve 0:21:19.200 --> 0:21:26.199 the jury's core function of making credibility determinations in criminal trials. 0:21:26.240 --> 0:21:29.040 In other words, they were saying, it should be about 0:21:29.080 --> 0:21:32.800 the person in the jury box making a decision, not 0:21:32.960 --> 0:21:38.280 about some machine or technology that says this is the truth. 0:21:38.920 --> 0:21:42.560 Now did their decision have something to do with the 0:21:42.880 --> 0:21:47.240 accuracy of the polygraph? Maybe? It turns out the polygraph 0:21:47.280 --> 0:21:51.400 test is actually quite good. In tests by the American 0:21:51.440 --> 0:21:55.600 Polygraph Association, they find it's about eighty seven percent accurate 0:21:55.840 --> 0:21:59.719 in determining whether someone is telling the truth or telling 0:21:59.760 --> 0:22:03.159 a although some scientists claim it's lower, maybe as low 0:22:03.200 --> 0:22:06.320 as seventy five percent. The thing to note is that's 0:22:06.359 --> 0:22:10.439 not bad, but some critics argue that you just can't 0:22:10.480 --> 0:22:14.199 accept and errorrate that high if somebody is guilt or 0:22:14.280 --> 0:22:18.480 innocence depends on this, and some courts argue that it's 0:22:18.480 --> 0:22:21.879 not just a matter of whether the accuracy is good enough. 0:22:22.359 --> 0:22:26.000 Part of the concern expressed by courts is whether jurors 0:22:26.560 --> 0:22:30.840 understand probabilities well enough. In other words, if a defendant 0:22:31.280 --> 0:22:35.440 fails the lie detector test, will a juror think, Okay, well, 0:22:35.520 --> 0:22:38.560 it's not one hundred percent accuracy. There's still a thirteen 0:22:38.600 --> 0:22:41.359 to twenty five percent chance that the machine got it wrong, 0:22:41.640 --> 0:22:44.440 and so here's how I'm going to work those probabilities 0:22:44.480 --> 0:22:49.359 into my decision. Many courts assumed, probably correctly, that the 0:22:49.440 --> 0:22:53.639 technology will sway people more than it should. They will 0:22:53.680 --> 0:22:58.679 believe whatever the machine says. And beyond the issue of 0:22:58.800 --> 0:23:02.080 jury interpretation, there's another problem as well, which is that 0:23:02.440 --> 0:23:07.320 some people are pathological liars. Some people, for example, those 0:23:07.359 --> 0:23:12.639 with psychopathy, have real deficits in empathy, and they're not 0:23:12.840 --> 0:23:15.960 good predictors of future punishment, and they lack what we 0:23:16.040 --> 0:23:20.240 might poetically call a conscience, and so for them, lying 0:23:20.680 --> 0:23:24.400 just doesn't trigger a stress response. Many of the studies 0:23:24.440 --> 0:23:28.720 on lie detection are done with undergraduates or otherwise normal people, 0:23:29.000 --> 0:23:32.760 but the people we really care about most are psychopaths, 0:23:32.840 --> 0:23:49.840 and the technology doesn't always work so well with them. 0:23:52.720 --> 0:23:57.080 So as it stands now, the polygraph test is generally 0:23:57.200 --> 0:23:59.960 not used at the federal level in the United States 0:24:00.600 --> 0:24:04.199 unless the prosecution and defense both agree to allow it, 0:24:04.680 --> 0:24:08.320 and at the state level it varies widely. Essentially one 0:24:08.480 --> 0:24:12.760 half of the states allow polygraph results while the other 0:24:12.840 --> 0:24:17.800 half generally prohibits them. But the polygraph test, which has 0:24:17.840 --> 0:24:21.040 been around for over a century now, is just one 0:24:21.160 --> 0:24:25.760 method that people use, and as science cranks forward, there 0:24:25.760 --> 0:24:29.360 have been many newer proposals. Actually a lot of these 0:24:29.400 --> 0:24:33.040 ideas sprung up after nine to eleven when people wondered 0:24:33.240 --> 0:24:36.600 how can you do lie detection on big groups of 0:24:36.640 --> 0:24:41.320 people quickly. So one of these techniques is thermal imaging. 0:24:41.359 --> 0:24:44.679 There's some evidence that when you're lying, you get hotter 0:24:44.880 --> 0:24:47.399 under the orbits of your eyes. So the idea is 0:24:47.400 --> 0:24:51.560 to use cameras at airports to do infrared imaging as 0:24:51.600 --> 0:24:55.520 people are passing through and answering questions. And people have 0:24:55.600 --> 0:24:59.320 also been looking at the subtle detection of stress in 0:24:59.640 --> 0:25:03.680 the voi, the way that your voice gets constricted when 0:25:03.720 --> 0:25:08.040 you have stress overlying about something. But things get a 0:25:08.080 --> 0:25:12.480 little more interesting and spicy when we start looking directly 0:25:12.640 --> 0:25:16.280 at signals of the brain, and one of these technologies 0:25:16.680 --> 0:25:20.159 involves sticking electrodes on the outside of the scalp, and 0:25:20.200 --> 0:25:24.119 this is known as electron cepholography or EEG. And the 0:25:24.320 --> 0:25:27.960 idea here is to use things like the guilty knowledge 0:25:28.080 --> 0:25:32.400 test to see if there's a big brain response when 0:25:32.480 --> 0:25:36.280 the person sees something they are familiar with. So let's 0:25:36.320 --> 0:25:38.960 go back to that story that I began with about 0:25:39.040 --> 0:25:43.960 Aditi Sharma and her boyfriend Pravine, who together killed her fiance. 0:25:45.080 --> 0:25:47.960 In Aditi's court trial in two thousand and eight, she 0:25:48.119 --> 0:25:50.840 was tested with a thirty two electrode system on her 0:25:50.880 --> 0:25:55.399 head called the Brain Electrical Oscillations Signature system. And the 0:25:55.440 --> 0:25:59.119 way it worked is this, a forensic scientist had a 0:25:59.200 --> 0:26:03.520 dit listen to a voice talk about the alleged murder 0:26:04.040 --> 0:26:08.320 and when she heard specific incriminating bits of the story 0:26:08.359 --> 0:26:12.200 about how the murder was committed. For example, he was 0:26:12.280 --> 0:26:15.000 killed with a special kind of drink in McDonald's that 0:26:15.080 --> 0:26:17.919 was poisoned with arsenic When she heard those bits of 0:26:17.920 --> 0:26:22.840 the story, woosh, her brain had larger activity, which was 0:26:22.880 --> 0:26:26.879